This document summarizes different approaches to tempo detection and beat tracking from audio signals, including oscillator, filterbank, and template-based approaches. The oscillator approach models beat tracking as an adaptive oscillator that predicts beat locations. The filterbank approach uses a bank of filters spaced at different intervals to detect the tempo. The template-based approach correlates templates of pulses at different tempos with the audio signal. Challenges discussed include detecting multiple hierarchical tempo levels and handling sudden tempo changes.
This document discusses onset detection in audio signals. It describes onset detection as a two-step process: 1) generating a novelty function to measure signal changes over time, and 2) peak picking to identify onset locations in the novelty function. Common novelty functions are based on time-domain envelopes, pitch contours, or STFT differences. Peak picking typically uses thresholding to find local maxima above a likelihood threshold. The goal of evaluation is to compare predicted onset times to ground truth labels.
This document presents a major project on hierarchical timing analysis of VLSI circuits. It includes an outline covering introduction, why timing analysis is needed, basics of timing analysis, static timing analysis, timing paths, hierarchical timing analysis applications, and conclusions. The introduction discusses using static timing analysis to ensure correct timing of clocks and signals. It also explains how hierarchical timing analysis can help alleviate large runtimes from flat analysis of growing design sizes. The document then covers various topics related to timing analysis including digital circuit to timing model conversions, static timing analysis concepts, different path types, and applications of hierarchical timing analysis.
The document discusses the beat histogram, which is a representation of the periodicities in an audio signal. It can be computed from a novelty function or series of onset times. Features are then extracted from the beat histogram like the mean, max peak value, and standard deviation to characterize the rhythmic properties. However, the beat histogram lacks phase information so it provides a limited description of rhythm.
This document is a group project report from the University of Sciences and Technologies of Hanoi that evaluates pitch detection algorithms and their application in musical key detection. The report was produced by a group of 5 students and their supervisor. The report provides an overview of pitch detection algorithms, including YIN estimation, cepstrum analysis, and simplified inverse filter tracking. It also covers musical key detection using pitch class profiles. The report describes the group's research process, implementation of the pitch detection algorithms in Java, and development of an Android application for musical key detection. It presents and discusses the group's results from testing the pitch detection algorithms and evaluating musical key detection.
2° Presentazione del workshop finale del progetto Step-by-Step
Obiettivo 2 - Fase Cronica: RIABILITAZIONE “EVIDENCE-BASED”
Lo sviluppo e la validazione clinica di un sistema di misura in grado di offrire una valutazione del percorso riabilitativo basata sull’evidenza, integrando le scale cliniche esistenti con parametri affidabili e ripetibili di segnali di movimento, correlabili con i valori delle scale.
This document provides an overview of techniques for fundamental frequency detection in polyphonic signals, including iterative subtraction methods and non-negative matrix factorization (NMF). Iterative subtraction methods work by repeatedly detecting the most salient frequency, removing its contribution, and iterating until all frequencies are identified. NMF decomposes a signal into additive combinations of spectral templates to separate voices. The document discusses algorithms like Cheveigné, Meddis, and Klapuri that apply these techniques, and compares their approaches, advantages, and limitations.
Combining Cluster Sampling and ACE analysis to improve fault-injection based ...Stefano Di Carlo
This document proposes combining cluster sampling and ACE analysis to improve the efficiency of fault injection-based reliability evaluation of GPU-based systems. It presents a methodology that uses ACE analysis to prune non-vulnerable faults, reducing the number that must be injected. It then applies a two-stage cluster sampling approach to further reduce simulation time while maintaining accuracy. Experimental results on a GPU microarchitectural simulator show the methodology reduces the number of required fault injections by over 800x on average compared to exhaustive fault injection.
This document discusses onset detection in audio signals. It describes onset detection as a two-step process: 1) generating a novelty function to measure signal changes over time, and 2) peak picking to identify onset locations in the novelty function. Common novelty functions are based on time-domain envelopes, pitch contours, or STFT differences. Peak picking typically uses thresholding to find local maxima above a likelihood threshold. The goal of evaluation is to compare predicted onset times to ground truth labels.
This document presents a major project on hierarchical timing analysis of VLSI circuits. It includes an outline covering introduction, why timing analysis is needed, basics of timing analysis, static timing analysis, timing paths, hierarchical timing analysis applications, and conclusions. The introduction discusses using static timing analysis to ensure correct timing of clocks and signals. It also explains how hierarchical timing analysis can help alleviate large runtimes from flat analysis of growing design sizes. The document then covers various topics related to timing analysis including digital circuit to timing model conversions, static timing analysis concepts, different path types, and applications of hierarchical timing analysis.
The document discusses the beat histogram, which is a representation of the periodicities in an audio signal. It can be computed from a novelty function or series of onset times. Features are then extracted from the beat histogram like the mean, max peak value, and standard deviation to characterize the rhythmic properties. However, the beat histogram lacks phase information so it provides a limited description of rhythm.
This document is a group project report from the University of Sciences and Technologies of Hanoi that evaluates pitch detection algorithms and their application in musical key detection. The report was produced by a group of 5 students and their supervisor. The report provides an overview of pitch detection algorithms, including YIN estimation, cepstrum analysis, and simplified inverse filter tracking. It also covers musical key detection using pitch class profiles. The report describes the group's research process, implementation of the pitch detection algorithms in Java, and development of an Android application for musical key detection. It presents and discusses the group's results from testing the pitch detection algorithms and evaluating musical key detection.
2° Presentazione del workshop finale del progetto Step-by-Step
Obiettivo 2 - Fase Cronica: RIABILITAZIONE “EVIDENCE-BASED”
Lo sviluppo e la validazione clinica di un sistema di misura in grado di offrire una valutazione del percorso riabilitativo basata sull’evidenza, integrando le scale cliniche esistenti con parametri affidabili e ripetibili di segnali di movimento, correlabili con i valori delle scale.
This document provides an overview of techniques for fundamental frequency detection in polyphonic signals, including iterative subtraction methods and non-negative matrix factorization (NMF). Iterative subtraction methods work by repeatedly detecting the most salient frequency, removing its contribution, and iterating until all frequencies are identified. NMF decomposes a signal into additive combinations of spectral templates to separate voices. The document discusses algorithms like Cheveigné, Meddis, and Klapuri that apply these techniques, and compares their approaches, advantages, and limitations.
Combining Cluster Sampling and ACE analysis to improve fault-injection based ...Stefano Di Carlo
This document proposes combining cluster sampling and ACE analysis to improve the efficiency of fault injection-based reliability evaluation of GPU-based systems. It presents a methodology that uses ACE analysis to prune non-vulnerable faults, reducing the number that must be injected. It then applies a two-stage cluster sampling approach to further reduce simulation time while maintaining accuracy. Experimental results on a GPU microarchitectural simulator show the methodology reduces the number of required fault injections by over 800x on average compared to exhaustive fault injection.
This document summarizes research into system identification and model validation for wave energy converters (WECs). It discusses using numerical modeling and experimental testing to develop and compare linear and nonlinear models of a WEC's dynamics. Both parametric and non-parametric models are considered, including models relating actuator force and wave elevation to velocity. The research aims to support more advanced WEC control design by providing quantitative models and analyses.
The document summarizes a seminar project on real-time recognition of ambulation mode in transfemoral prostheses. Sensors including IMUs and pressure sensors will be used to collect locomotion data from different ambulation modes. An ESP 32 device will transmit the sensor data, which will be stored in Excel. IMUs will be placed on the thigh, shank, and foot to measure acceleration and gyroscopic data. Pressure sensors on the heel and toes will also record data. Various machine learning algorithms will be tested and compared to accurately classify ambulation modes such as walking on flat ground versus up/down ramps or stairs.
Isa saint-louis-exceptional-opportunities-short-course-day-1Jim Cahill
The document provides an overview of a three-day process control conference in December 2010. Day 1 will include a welcome from keynote speaker Greg McMillan, who has extensive experience in process control. The rest of Day 1 will focus on common mistakes in project definition meetings and the top 10 concepts in process control, including loop deadtime, speed, gain, resonance, attenuation, sensitivity and resolution.
This document provides details about the course "Sensors and Transducers" offered at SRM Institute of Science & Technology. It includes information about the course code, category, credits, pre-requisites, learning objectives, expected outcomes, syllabus structure over 9 sessions, assessment details and course designers. The course aims to provide knowledge about different types of sensors for measurement of various electrical and non-electrical quantities and their applications in industries and home appliances. Students will learn about transducers, inductive, capacitive, thermal, magnetic and other sensors through theory and practical sessions. Continuous and final assessments include tests, assignments and projects to evaluate students' understanding of sensor fundamentals and applications.
The document discusses molecular clock methods for estimating divergence times between species. It provides an overview of molecular clocks and describes several software packages used for dating species divergences, including BEAST, MCMCTree, MultiDivTime, and others. It also discusses important considerations for molecular clock analyses such as calibration priors, clock models, evaluating convergence, and interpreting credibility intervals.
Favorite Delay Analysis Methodologies Town Hall SEIChris Carson
Presentation from a Town Hall session to discuss favorite forensic schedule analysis methodologies, based on the Forensic Analysis Recommended Practice from AACE International. The Best Practices and Guidelines for Schedule Impact Analysis project is discussing methods.
The document provides guidance on using random sequential sampling with Excel and Audit Commander software. It describes sampling population data in Excel worksheets and selecting initial samples of 150 observations for variable and attribute sampling. The samples are sorted by random numbers and values from the actual column are copied over to the audited value column.
Specification and Detection of SOA AntipatternsFrancis Palma
SODA is a novel approach for specifying and detecting SOA antipatterns. It involves specifying antipatterns using a domain-specific language, generating detection algorithms from the specifications, and applying the algorithms to detect antipatterns in service-based systems. An evaluation on two versions of a home automation system found that SODA could accurately detect 10 different antipatterns, with a recall of 100% and precision over 75%. The approach demonstrates potential for automated analysis and improvement of service-oriented architectures.
Essentials of jitter part 1 The Time Interval Error: TIEteledynelecroy
This document discusses jitter and its measurement. Jitter is measured as the time interval error (TIE) between the actual and expected arrival times of signal edges. TIE values over time form the TIE track waveform, which can be analyzed statistically, in the frequency domain, and through histograms to identify different jitter sources. Proper jitter measurement requires determining the expected edge times, which may involve clock data recovery for signals without a reference clock.
The experiment measured the firing rate of neurons in the ventral nerve cord of an anesthetized cockroach when exposed to different sound frequencies, including 200Hz, 400Hz, 600Hz, and an ultrasonic pest repellent of 25,000Hz. Eight trials were conducted for each frequency, and the average firing rate was calculated and statistically analyzed to determine the effect of different sound frequencies on neural activity. Electrodes were used to record extracellular action potentials from neurons and an oscilloscope and statistical tests analyzed the results.
Making & Breaking Machine Learning Anomaly Detectors in Real Life by Clarence...CODE BLUE
Machine learning-based (ML) techniques for network intrusion detection have gained notable traction in the web security industry over the past decade. Some Intrusion Detection Systems (IDS) successfully used these techniques to detect and deflect network intrusions before they could cause significant harm to network services. Simply put, IDS systems construct a signature model of how normal traffic looks, using data retrieved from web access logs as input. Then, an online processing system is put in place to maintain a model of how expected network traffic looks like, and/or how malicious traffic looks like. When traffic that is deviant from the expected model exceeds the defined threshold, the IDS flags it as malicious. The theory behind it was that the more data the system sees, the more accurate the model would become. This provides a flexible system for traffic analysis, seemingly perfect for the constantly evolving and growing web traffic patterns.
However, this fairytale did not last for long. It was soon found that the attackers had been avoiding detection by ‘poisoning’ the classifier models used by these PCA systems. The adversaries slowly train the detection model by sending large volumes of seemingly benign web traffic to make the classification model more tolerant to outliers and actual malicious attempts. They succeeded.
In this talk, we will do a live demo of this 'model-poisoning' attack and analyze methods that have been proposed to decrease the susceptibility of ML-based network anomaly detection systems from being manipulated by attackers. Instead of diving into the ML theory behind this, we will emphasize on examples of these systems working in the real world, the attacks that render them impotent, and how it affects developers looking to protect themselves from network intrusion. Most importantly, we will look towards the future of ML-based network intrusion detection.
Analysis for Radar and Electronic WarfareReza Taryghat
This document discusses techniques for measuring pulsed RF signals used in radar and electronic warfare applications. It begins with an overview of common radar applications and measurement types. It then discusses tools for measuring pulse parameters like pulse width, repetition interval, and power. These tools include power meters, oscilloscopes, spectrum analyzers, and specialized pulse analyzers. It also covers vector signal analysis and its ability to analyze modulation embedded on pulses. The rest of the document provides examples of measuring pulses with these various tools and techniques like pulse building, frequency hopping analysis, and analyzing LFM chirps.
From gamma-ray to radio: Multi-wavelength follow-up in the first five minutesTim Staley
This document summarizes recent work on fast radio follow-up of transient sources and multi-wavelength classification. It discusses three key areas: 1) examples of fast radio follow-up of gamma-ray bursts and stellar flares, 2) using radio and optical measurements to classify transients, and 3) the need to automate "transient triage" to efficiently prioritize follow-up observations across different facilities. The presenter argues that distributing information about new transients via VOEvents can help the community automate real-time classification and prioritization of targets.
The document describes the Apex Instruments Model XC-500 Series isokinetic source sampling system. It provides an overview of the key components of the system including the source sampler console, electrical subsystem, thermocouple subsystem, vacuum subsystem, external vacuum pump unit, probe assembly, modular sample case, umbilical cable, and glassware sample train. It also describes the operating procedures, test design, site preparation, sampling methods, calibration procedures, and maintenance for the system. The system is applicable for performing isokinetic testing and sampling for various pollutants according to EPA reference methods.
This document provides an introduction to the fundamentals of digitizing audio content, including discretization of signals through sampling in time and quantization in amplitude. It describes key concepts such as the sampling theorem, which states that a sampled signal can be perfectly reconstructed if the sampling frequency is greater than twice the maximum frequency of the original signal. It also covers properties of the quantization error introduced by representing continuous amplitude values with a finite number of levels, and how factors like word length affect the signal-to-noise ratio of the quantization process.
This thesis presents an observer-based method for parameter estimation and its application to model-based fault detection, isolation, and estimation. The method uses two coupled Luenberger observers with a delay to estimate states in finite time, allowing consistent parameter estimation from a single data point if the estimated states meet a full rank condition. The estimated parameters are then used in an Information Synthesis approach for one-step-ahead fault detection, isolation, and estimation. The thesis validates the observer-based parameter estimation method on a numerical example of a double mass-spring-damper system and experimental sensor fault diagnosis data from an engine.
6.5.1 Review of Beat Detection and Phase Alignment Techniques used in DMMNasir Ahmad
The document describes a beat detection and phase alignment algorithm used in a music morphing application. It analyzes the signal in multiple frequency bands using onset detection and autocorrelation. Tempo is extracted by feeding the autocorrelation outputs into a bank of pulse trains tuned to different tempos. The pulse train that phase locks to the signal indicates the tempo. Phase is determined by convolving the onset detection with a resonator comb filter tuned to the extracted tempo. The algorithm was able to reliably detect tempo within 3 BPM and phase align signals for seamless morphing.
Estimating instruction-level throughput (for example, predicting the cycle counts) is critical for many applications that rely on tightly calculated and accurate timing bounds. In this talk, we will present a new throughput analysis tool, MCA Daemon (MCAD). It is built on top of LLVM MCA and combines the advantages of both static and dynamic throughput analyses, providing a powerful, fast, and easy-to-use tool that scales up with large-scale programs in the real world.
This document summarizes Kiran Lakhotia's PhD thesis on search-based testing. The thesis aims to advance automated test data generation using search-based techniques by addressing challenges like pointers, dynamic data structures, and loop-assigned flags. It presents a tool called AUSTIN for generating test data for C programs. AUSTIN combines a hill climber search algorithm with a custom constraint solver for pointers. The thesis also introduces a testability transformation to handle loop-assigned flags and a multi-objective approach to optimize for branch coverage and memory usage. Empirical studies compare AUSTIN to other tools on large open source programs and embedded software.
This document provides an introduction to audio content analysis and convolution. It discusses linear time-invariant (LTI) systems and how convolution describes the output of an LTI system based on the input and impulse response. Examples of simple low-pass filters like moving average and single-pole filters are provided to illustrate convolution. The objectives are to understand LTI systems, convolution operations, and implement basic filters.
This document provides an overview of musical genre classification. It discusses how genre classification is one of the early research topics in music information retrieval and involves extracting audio features from songs and using those features to classify the songs into genres. The document describes common features used like spectral, pitch, rhythm, and intensity features. It also outlines the typical processing steps of feature extraction, normalization, selection and classification. Examples of challenges with genre classification like ill-defined genres and overfitting are also mentioned.
This document summarizes research into system identification and model validation for wave energy converters (WECs). It discusses using numerical modeling and experimental testing to develop and compare linear and nonlinear models of a WEC's dynamics. Both parametric and non-parametric models are considered, including models relating actuator force and wave elevation to velocity. The research aims to support more advanced WEC control design by providing quantitative models and analyses.
The document summarizes a seminar project on real-time recognition of ambulation mode in transfemoral prostheses. Sensors including IMUs and pressure sensors will be used to collect locomotion data from different ambulation modes. An ESP 32 device will transmit the sensor data, which will be stored in Excel. IMUs will be placed on the thigh, shank, and foot to measure acceleration and gyroscopic data. Pressure sensors on the heel and toes will also record data. Various machine learning algorithms will be tested and compared to accurately classify ambulation modes such as walking on flat ground versus up/down ramps or stairs.
Isa saint-louis-exceptional-opportunities-short-course-day-1Jim Cahill
The document provides an overview of a three-day process control conference in December 2010. Day 1 will include a welcome from keynote speaker Greg McMillan, who has extensive experience in process control. The rest of Day 1 will focus on common mistakes in project definition meetings and the top 10 concepts in process control, including loop deadtime, speed, gain, resonance, attenuation, sensitivity and resolution.
This document provides details about the course "Sensors and Transducers" offered at SRM Institute of Science & Technology. It includes information about the course code, category, credits, pre-requisites, learning objectives, expected outcomes, syllabus structure over 9 sessions, assessment details and course designers. The course aims to provide knowledge about different types of sensors for measurement of various electrical and non-electrical quantities and their applications in industries and home appliances. Students will learn about transducers, inductive, capacitive, thermal, magnetic and other sensors through theory and practical sessions. Continuous and final assessments include tests, assignments and projects to evaluate students' understanding of sensor fundamentals and applications.
The document discusses molecular clock methods for estimating divergence times between species. It provides an overview of molecular clocks and describes several software packages used for dating species divergences, including BEAST, MCMCTree, MultiDivTime, and others. It also discusses important considerations for molecular clock analyses such as calibration priors, clock models, evaluating convergence, and interpreting credibility intervals.
Favorite Delay Analysis Methodologies Town Hall SEIChris Carson
Presentation from a Town Hall session to discuss favorite forensic schedule analysis methodologies, based on the Forensic Analysis Recommended Practice from AACE International. The Best Practices and Guidelines for Schedule Impact Analysis project is discussing methods.
The document provides guidance on using random sequential sampling with Excel and Audit Commander software. It describes sampling population data in Excel worksheets and selecting initial samples of 150 observations for variable and attribute sampling. The samples are sorted by random numbers and values from the actual column are copied over to the audited value column.
Specification and Detection of SOA AntipatternsFrancis Palma
SODA is a novel approach for specifying and detecting SOA antipatterns. It involves specifying antipatterns using a domain-specific language, generating detection algorithms from the specifications, and applying the algorithms to detect antipatterns in service-based systems. An evaluation on two versions of a home automation system found that SODA could accurately detect 10 different antipatterns, with a recall of 100% and precision over 75%. The approach demonstrates potential for automated analysis and improvement of service-oriented architectures.
Essentials of jitter part 1 The Time Interval Error: TIEteledynelecroy
This document discusses jitter and its measurement. Jitter is measured as the time interval error (TIE) between the actual and expected arrival times of signal edges. TIE values over time form the TIE track waveform, which can be analyzed statistically, in the frequency domain, and through histograms to identify different jitter sources. Proper jitter measurement requires determining the expected edge times, which may involve clock data recovery for signals without a reference clock.
The experiment measured the firing rate of neurons in the ventral nerve cord of an anesthetized cockroach when exposed to different sound frequencies, including 200Hz, 400Hz, 600Hz, and an ultrasonic pest repellent of 25,000Hz. Eight trials were conducted for each frequency, and the average firing rate was calculated and statistically analyzed to determine the effect of different sound frequencies on neural activity. Electrodes were used to record extracellular action potentials from neurons and an oscilloscope and statistical tests analyzed the results.
Making & Breaking Machine Learning Anomaly Detectors in Real Life by Clarence...CODE BLUE
Machine learning-based (ML) techniques for network intrusion detection have gained notable traction in the web security industry over the past decade. Some Intrusion Detection Systems (IDS) successfully used these techniques to detect and deflect network intrusions before they could cause significant harm to network services. Simply put, IDS systems construct a signature model of how normal traffic looks, using data retrieved from web access logs as input. Then, an online processing system is put in place to maintain a model of how expected network traffic looks like, and/or how malicious traffic looks like. When traffic that is deviant from the expected model exceeds the defined threshold, the IDS flags it as malicious. The theory behind it was that the more data the system sees, the more accurate the model would become. This provides a flexible system for traffic analysis, seemingly perfect for the constantly evolving and growing web traffic patterns.
However, this fairytale did not last for long. It was soon found that the attackers had been avoiding detection by ‘poisoning’ the classifier models used by these PCA systems. The adversaries slowly train the detection model by sending large volumes of seemingly benign web traffic to make the classification model more tolerant to outliers and actual malicious attempts. They succeeded.
In this talk, we will do a live demo of this 'model-poisoning' attack and analyze methods that have been proposed to decrease the susceptibility of ML-based network anomaly detection systems from being manipulated by attackers. Instead of diving into the ML theory behind this, we will emphasize on examples of these systems working in the real world, the attacks that render them impotent, and how it affects developers looking to protect themselves from network intrusion. Most importantly, we will look towards the future of ML-based network intrusion detection.
Analysis for Radar and Electronic WarfareReza Taryghat
This document discusses techniques for measuring pulsed RF signals used in radar and electronic warfare applications. It begins with an overview of common radar applications and measurement types. It then discusses tools for measuring pulse parameters like pulse width, repetition interval, and power. These tools include power meters, oscilloscopes, spectrum analyzers, and specialized pulse analyzers. It also covers vector signal analysis and its ability to analyze modulation embedded on pulses. The rest of the document provides examples of measuring pulses with these various tools and techniques like pulse building, frequency hopping analysis, and analyzing LFM chirps.
From gamma-ray to radio: Multi-wavelength follow-up in the first five minutesTim Staley
This document summarizes recent work on fast radio follow-up of transient sources and multi-wavelength classification. It discusses three key areas: 1) examples of fast radio follow-up of gamma-ray bursts and stellar flares, 2) using radio and optical measurements to classify transients, and 3) the need to automate "transient triage" to efficiently prioritize follow-up observations across different facilities. The presenter argues that distributing information about new transients via VOEvents can help the community automate real-time classification and prioritization of targets.
The document describes the Apex Instruments Model XC-500 Series isokinetic source sampling system. It provides an overview of the key components of the system including the source sampler console, electrical subsystem, thermocouple subsystem, vacuum subsystem, external vacuum pump unit, probe assembly, modular sample case, umbilical cable, and glassware sample train. It also describes the operating procedures, test design, site preparation, sampling methods, calibration procedures, and maintenance for the system. The system is applicable for performing isokinetic testing and sampling for various pollutants according to EPA reference methods.
This document provides an introduction to the fundamentals of digitizing audio content, including discretization of signals through sampling in time and quantization in amplitude. It describes key concepts such as the sampling theorem, which states that a sampled signal can be perfectly reconstructed if the sampling frequency is greater than twice the maximum frequency of the original signal. It also covers properties of the quantization error introduced by representing continuous amplitude values with a finite number of levels, and how factors like word length affect the signal-to-noise ratio of the quantization process.
This thesis presents an observer-based method for parameter estimation and its application to model-based fault detection, isolation, and estimation. The method uses two coupled Luenberger observers with a delay to estimate states in finite time, allowing consistent parameter estimation from a single data point if the estimated states meet a full rank condition. The estimated parameters are then used in an Information Synthesis approach for one-step-ahead fault detection, isolation, and estimation. The thesis validates the observer-based parameter estimation method on a numerical example of a double mass-spring-damper system and experimental sensor fault diagnosis data from an engine.
6.5.1 Review of Beat Detection and Phase Alignment Techniques used in DMMNasir Ahmad
The document describes a beat detection and phase alignment algorithm used in a music morphing application. It analyzes the signal in multiple frequency bands using onset detection and autocorrelation. Tempo is extracted by feeding the autocorrelation outputs into a bank of pulse trains tuned to different tempos. The pulse train that phase locks to the signal indicates the tempo. Phase is determined by convolving the onset detection with a resonator comb filter tuned to the extracted tempo. The algorithm was able to reliably detect tempo within 3 BPM and phase align signals for seamless morphing.
Estimating instruction-level throughput (for example, predicting the cycle counts) is critical for many applications that rely on tightly calculated and accurate timing bounds. In this talk, we will present a new throughput analysis tool, MCA Daemon (MCAD). It is built on top of LLVM MCA and combines the advantages of both static and dynamic throughput analyses, providing a powerful, fast, and easy-to-use tool that scales up with large-scale programs in the real world.
This document summarizes Kiran Lakhotia's PhD thesis on search-based testing. The thesis aims to advance automated test data generation using search-based techniques by addressing challenges like pointers, dynamic data structures, and loop-assigned flags. It presents a tool called AUSTIN for generating test data for C programs. AUSTIN combines a hill climber search algorithm with a custom constraint solver for pointers. The thesis also introduces a testability transformation to handle loop-assigned flags and a multi-objective approach to optimize for branch coverage and memory usage. Empirical studies compare AUSTIN to other tools on large open source programs and embedded software.
This document provides an introduction to audio content analysis and convolution. It discusses linear time-invariant (LTI) systems and how convolution describes the output of an LTI system based on the input and impulse response. Examples of simple low-pass filters like moving average and single-pole filters are provided to illustrate convolution. The objectives are to understand LTI systems, convolution operations, and implement basic filters.
This document provides an overview of musical genre classification. It discusses how genre classification is one of the early research topics in music information retrieval and involves extracting audio features from songs and using those features to classify the songs into genres. The document describes common features used like spectral, pitch, rhythm, and intensity features. It also outlines the typical processing steps of feature extraction, normalization, selection and classification. Examples of challenges with genre classification like ill-defined genres and overfitting are also mentioned.
This document summarizes an introduction to audio content analysis module on audio-to-audio and audio-to-score alignment. It discusses using audio feature extraction and dynamic time warping to align two audio sequences or an audio sequence to a musical score. Key steps include extracting features from the audio, computing a distance matrix using measures like Euclidean distance, and finding the optimal alignment path. Evaluation of alignment accuracy can be challenging due to differences in the domains of audio and scores.
This document provides an overview of mood recognition in audio content analysis. It discusses how mood recognition aims to identify the mood or emotion of a song by extracting features and classifying using methods like regression. Key challenges include defining and quantifying moods/emotions, developing ground truth data, and determining appropriate mood models. Common approaches include classifying into mood clusters or mapping to the arousal-valence circumplex model of affect using regression techniques. Results can vary significantly based on the data but classification rates around 50% and regression errors of 0.1-0.4 are typical.
This document discusses music structure detection. It introduces self-similarity matrices which show pairwise similarities between segments of a song and can be used to detect structural changes based on novelty, homogeneity, or repetitions. Ground truth annotations of music structure are challenging due to ambiguity and different hierarchical levels. Typical accuracy of automatic structure detection ranges from 50-70% compared to human annotations.
This document introduces key concepts for analyzing musical rhythm, including onset, tempo, meter, bar, and rhythm. It discusses how humans perceive and group temporal musical events, and describes typical onset durations for different instruments. Accuracy for detecting onsets ranges from 2-300ms, depending on factors like tempo and instrumentation. Musical works represent these temporal concepts through notation of tempo, time signature, bars, and note values.
This document discusses chord detection in music audio. It introduces musical chords and their properties, describes the process of extracting pitch chroma and comparing it to chord templates, and explains how Hidden Markov Models (HMMs) can be used to model chord progressions over time using transition probabilities between states representing different chords. The Viterbi algorithm is used to find the most likely chord sequence given the observations.
This document provides an overview of music performance analysis, including defining music performance, common parameters analyzed like tempo and dynamics, goals of performance analysis like understanding the performer and listener experience, challenges like disentangling variables and limited reliable datasets, and opportunities to address challenges through cross-disciplinary work.
This document discusses music similarity analysis and clustering. It describes how music similarity is multidimensional and subjective. Common approaches for measuring music similarity involve extracting audio features and calculating distances between songs in the feature space. Visualization techniques like self-organizing maps can be used to project the high-dimensional feature spaces into 2D for visualization. Evaluation is challenging as there is no clear ground truth, but can involve using genre labels or listening surveys.
This document provides an overview of audio fingerprinting. It discusses the goals of representing an audio recording with a compact fingerprint and matching fingerprints to identify recordings. It describes fingerprinting applications in broadcast monitoring and value-added services. The document outlines fingerprint requirements and compares fingerprinting to watermarking. It then details the processing steps in Philips' audio fingerprinting system, including fingerprint extraction, database storage, and identification.
This document provides an overview of an audio content analysis module about intensity and loudness. It discusses human perception of loudness and how it relates to physical measurements of intensity and magnitude. It then describes several common low-level features used to represent intensity in audio signals, including root mean square (RMS), peak envelope, weighted RMS, and models of loudness like the Zwicker model. The goal is to discuss intensity and loudness from both a perceptual and technical feature extraction perspective.
This document provides an overview of instantaneous audio features that can be extracted from audio signals. It discusses that features are low-level descriptors of audio that are not necessarily perceptually meaningful on their own. Common instantaneous features described include spectral centroid, spectral spread, spectral rolloff, and spectral flux, which characterize the spectral shape and are useful for timbre perception. Mel-frequency cepstral coefficients (MFCCs) are also covered as they represent the short-term power spectrum of a sound. The document aims to describe the computation and meaning of various instantaneous audio features.
This document discusses approaches for detecting the fundamental frequency (f0) of monophonic audio signals. It describes time-domain methods like zero-crossing rate analysis and autocorrelation, as well as frequency-domain techniques like harmonic product spectrum, harmonic sum spectrum, and spectral maximum likelihood. It also covers auditory-motivated approaches and notes that while older, tweaked versions of these methods remain state-of-the-art for monophonic f0 detection.
This document discusses data requirements, data splits, cross validation, and data augmentation for machine learning. It covers dividing data into training, validation, and test sets; using cross validation to evaluate models; and techniques for augmenting data like degrading quality, adding effects, and segmentation when data is insufficient. The overall goal is to have representative, clean data that is sufficient for the task and to maximize its use through validation and augmentation.
This document discusses evaluating pitch tracking systems by comparing predicted pitch to ground truth pitch. It describes challenges in annotating pitch and time, and defines different pitch tracking tasks from individual notes to polyphonic mixtures. Metrics for score-based and F0-based ground truths are provided to measure accuracy and deviation between predicted and ground truth pitch. Results should be aggregated at the frame/note and file levels.
This document provides an introduction and overview of time-frequency representations using the Fourier transform. It defines the continuous Fourier transform and discusses its key properties including invertibility, superposition, convolution/multiplication relationships, Parseval's theorem, and effects of time/frequency shifts and scaling. It also covers the Fourier transform of sampled signals, the short-time Fourier transform (STFT), and the discrete Fourier transform (DFT).
This document discusses feature learning techniques for audio content analysis. It introduces feature learning as an alternative to traditional hand-crafted features that can automatically learn features from data. Two main approaches are dictionary learning, which learns a dictionary of templates to represent audio signals, and deep learning using neural networks to learn high-level representations through multiple layers. Learned features may contain more useful information than hand-crafted ones but require large amounts of data and computation.
This document discusses post-processing of audio content features. It describes deriving additional features, aggregating features to reduce data and compute summary statistics, and normalizing features to ensure similar scales and distributions. Specific techniques include taking differences between adjacent values to derive new features, smoothing features using filters, computing z-score and min-max normalization, and aggregating features within blocks using statistical descriptors. The overall aim is to adapt the feature matrix to the task and classifier.
This document discusses the representation of pitch in music. It introduces musical pitch intervals and how they are notated, explains how MIDI represents pitch as a numeric value, defines cents as a measure of pitch distance, and discusses temperament, intonation, and vibrato in musical pitch performance and perception.
This document discusses evaluation metrics for machine learning models. It covers classification metrics like accuracy, precision, recall and F1 score. Regression metrics like MAE, MSE and R2 are also discussed. The document stresses the importance of proper evaluation methodology, including using representative test sets, comparing to baselines, and testing for statistical significance. Good practices outlined include task-driven not algorithm-driven evaluation and using established datasets and metrics.
Observability Concepts EVERY Developer Should Know -- DeveloperWeek Europe.pdfPaige Cruz
Monitoring and observability aren’t traditionally found in software curriculums and many of us cobble this knowledge together from whatever vendor or ecosystem we were first introduced to and whatever is a part of your current company’s observability stack.
While the dev and ops silo continues to crumble….many organizations still relegate monitoring & observability as the purview of ops, infra and SRE teams. This is a mistake - achieving a highly observable system requires collaboration up and down the stack.
I, a former op, would like to extend an invitation to all application developers to join the observability party will share these foundational concepts to build on:
Infrastructure Challenges in Scaling RAG with Custom AI modelsZilliz
Building Retrieval-Augmented Generation (RAG) systems with open-source and custom AI models is a complex task. This talk explores the challenges in productionizing RAG systems, including retrieval performance, response synthesis, and evaluation. We’ll discuss how to leverage open-source models like text embeddings, language models, and custom fine-tuned models to enhance RAG performance. Additionally, we’ll cover how BentoML can help orchestrate and scale these AI components efficiently, ensuring seamless deployment and management of RAG systems in the cloud.
Driving Business Innovation: Latest Generative AI Advancements & Success StorySafe Software
Are you ready to revolutionize how you handle data? Join us for a webinar where we’ll bring you up to speed with the latest advancements in Generative AI technology and discover how leveraging FME with tools from giants like Google Gemini, Amazon, and Microsoft OpenAI can supercharge your workflow efficiency.
During the hour, we’ll take you through:
Guest Speaker Segment with Hannah Barrington: Dive into the world of dynamic real estate marketing with Hannah, the Marketing Manager at Workspace Group. Hear firsthand how their team generates engaging descriptions for thousands of office units by integrating diverse data sources—from PDF floorplans to web pages—using FME transformers, like OpenAIVisionConnector and AnthropicVisionConnector. This use case will show you how GenAI can streamline content creation for marketing across the board.
Ollama Use Case: Learn how Scenario Specialist Dmitri Bagh has utilized Ollama within FME to input data, create custom models, and enhance security protocols. This segment will include demos to illustrate the full capabilities of FME in AI-driven processes.
Custom AI Models: Discover how to leverage FME to build personalized AI models using your data. Whether it’s populating a model with local data for added security or integrating public AI tools, find out how FME facilitates a versatile and secure approach to AI.
We’ll wrap up with a live Q&A session where you can engage with our experts on your specific use cases, and learn more about optimizing your data workflows with AI.
This webinar is ideal for professionals seeking to harness the power of AI within their data management systems while ensuring high levels of customization and security. Whether you're a novice or an expert, gain actionable insights and strategies to elevate your data processes. Join us to see how FME and AI can revolutionize how you work with data!
Goodbye Windows 11: Make Way for Nitrux Linux 3.5.0!SOFTTECHHUB
As the digital landscape continually evolves, operating systems play a critical role in shaping user experiences and productivity. The launch of Nitrux Linux 3.5.0 marks a significant milestone, offering a robust alternative to traditional systems such as Windows 11. This article delves into the essence of Nitrux Linux 3.5.0, exploring its unique features, advantages, and how it stands as a compelling choice for both casual users and tech enthusiasts.
For the full video of this presentation, please visit: https://www.edge-ai-vision.com/2024/06/building-and-scaling-ai-applications-with-the-nx-ai-manager-a-presentation-from-network-optix/
Robin van Emden, Senior Director of Data Science at Network Optix, presents the “Building and Scaling AI Applications with the Nx AI Manager,” tutorial at the May 2024 Embedded Vision Summit.
In this presentation, van Emden covers the basics of scaling edge AI solutions using the Nx tool kit. He emphasizes the process of developing AI models and deploying them globally. He also showcases the conversion of AI models and the creation of effective edge AI pipelines, with a focus on pre-processing, model conversion, selecting the appropriate inference engine for the target hardware and post-processing.
van Emden shows how Nx can simplify the developer’s life and facilitate a rapid transition from concept to production-ready applications.He provides valuable insights into developing scalable and efficient edge AI solutions, with a strong focus on practical implementation.
Threats to mobile devices are more prevalent and increasing in scope and complexity. Users of mobile devices desire to take full advantage of the features
available on those devices, but many of the features provide convenience and capability but sacrifice security. This best practices guide outlines steps the users can take to better protect personal devices and information.
Cosa hanno in comune un mattoncino Lego e la backdoor XZ?Speck&Tech
ABSTRACT: A prima vista, un mattoncino Lego e la backdoor XZ potrebbero avere in comune il fatto di essere entrambi blocchi di costruzione, o dipendenze di progetti creativi e software. La realtà è che un mattoncino Lego e il caso della backdoor XZ hanno molto di più di tutto ciò in comune.
Partecipate alla presentazione per immergervi in una storia di interoperabilità, standard e formati aperti, per poi discutere del ruolo importante che i contributori hanno in una comunità open source sostenibile.
BIO: Sostenitrice del software libero e dei formati standard e aperti. È stata un membro attivo dei progetti Fedora e openSUSE e ha co-fondato l'Associazione LibreItalia dove è stata coinvolta in diversi eventi, migrazioni e formazione relativi a LibreOffice. In precedenza ha lavorato a migrazioni e corsi di formazione su LibreOffice per diverse amministrazioni pubbliche e privati. Da gennaio 2020 lavora in SUSE come Software Release Engineer per Uyuni e SUSE Manager e quando non segue la sua passione per i computer e per Geeko coltiva la sua curiosità per l'astronomia (da cui deriva il suo nickname deneb_alpha).
GraphSummit Singapore | The Art of the Possible with Graph - Q2 2024Neo4j
Neha Bajwa, Vice President of Product Marketing, Neo4j
Join us as we explore breakthrough innovations enabled by interconnected data and AI. Discover firsthand how organizations use relationships in data to uncover contextual insights and solve our most pressing challenges – from optimizing supply chains, detecting fraud, and improving customer experiences to accelerating drug discoveries.
UiPath Test Automation using UiPath Test Suite series, part 6DianaGray10
Welcome to UiPath Test Automation using UiPath Test Suite series part 6. In this session, we will cover Test Automation with generative AI and Open AI.
UiPath Test Automation with generative AI and Open AI webinar offers an in-depth exploration of leveraging cutting-edge technologies for test automation within the UiPath platform. Attendees will delve into the integration of generative AI, a test automation solution, with Open AI advanced natural language processing capabilities.
Throughout the session, participants will discover how this synergy empowers testers to automate repetitive tasks, enhance testing accuracy, and expedite the software testing life cycle. Topics covered include the seamless integration process, practical use cases, and the benefits of harnessing AI-driven automation for UiPath testing initiatives. By attending this webinar, testers, and automation professionals can gain valuable insights into harnessing the power of AI to optimize their test automation workflows within the UiPath ecosystem, ultimately driving efficiency and quality in software development processes.
What will you get from this session?
1. Insights into integrating generative AI.
2. Understanding how this integration enhances test automation within the UiPath platform
3. Practical demonstrations
4. Exploration of real-world use cases illustrating the benefits of AI-driven test automation for UiPath
Topics covered:
What is generative AI
Test Automation with generative AI and Open AI.
UiPath integration with generative AI
Speaker:
Deepak Rai, Automation Practice Lead, Boundaryless Group and UiPath MVP
Full-RAG: A modern architecture for hyper-personalizationZilliz
Mike Del Balso, CEO & Co-Founder at Tecton, presents "Full RAG," a novel approach to AI recommendation systems, aiming to push beyond the limitations of traditional models through a deep integration of contextual insights and real-time data, leveraging the Retrieval-Augmented Generation architecture. This talk will outline Full RAG's potential to significantly enhance personalization, address engineering challenges such as data management and model training, and introduce data enrichment with reranking as a key solution. Attendees will gain crucial insights into the importance of hyperpersonalization in AI, the capabilities of Full RAG for advanced personalization, and strategies for managing complex data integrations for deploying cutting-edge AI solutions.
Removing Uninteresting Bytes in Software FuzzingAftab Hussain
Imagine a world where software fuzzing, the process of mutating bytes in test seeds to uncover hidden and erroneous program behaviors, becomes faster and more effective. A lot depends on the initial seeds, which can significantly dictate the trajectory of a fuzzing campaign, particularly in terms of how long it takes to uncover interesting behaviour in your code. We introduce DIAR, a technique designed to speedup fuzzing campaigns by pinpointing and eliminating those uninteresting bytes in the seeds. Picture this: instead of wasting valuable resources on meaningless mutations in large, bloated seeds, DIAR removes the unnecessary bytes, streamlining the entire process.
In this work, we equipped AFL, a popular fuzzer, with DIAR and examined two critical Linux libraries -- Libxml's xmllint, a tool for parsing xml documents, and Binutil's readelf, an essential debugging and security analysis command-line tool used to display detailed information about ELF (Executable and Linkable Format). Our preliminary results show that AFL+DIAR does not only discover new paths more quickly but also achieves higher coverage overall. This work thus showcases how starting with lean and optimized seeds can lead to faster, more comprehensive fuzzing campaigns -- and DIAR helps you find such seeds.
- These are slides of the talk given at IEEE International Conference on Software Testing Verification and Validation Workshop, ICSTW 2022.
Dr. Sean Tan, Head of Data Science, Changi Airport Group
Discover how Changi Airport Group (CAG) leverages graph technologies and generative AI to revolutionize their search capabilities. This session delves into the unique search needs of CAG’s diverse passengers and customers, showcasing how graph data structures enhance the accuracy and relevance of AI-generated search results, mitigating the risk of “hallucinations” and improving the overall customer journey.
GraphSummit Singapore | The Future of Agility: Supercharging Digital Transfor...Neo4j
Leonard Jayamohan, Partner & Generative AI Lead, Deloitte
This keynote will reveal how Deloitte leverages Neo4j’s graph power for groundbreaking digital twin solutions, achieving a staggering 100x performance boost. Discover the essential role knowledge graphs play in successful generative AI implementations. Plus, get an exclusive look at an innovative Neo4j + Generative AI solution Deloitte is developing in-house.
Programming Foundation Models with DSPy - Meetup SlidesZilliz
Prompting language models is hard, while programming language models is easy. In this talk, I will discuss the state-of-the-art framework DSPy for programming foundation models with its powerful optimizers and runtime constraint system.
HCL Notes and Domino License Cost Reduction in the World of DLAUpanagenda
Webinar Recording: https://www.panagenda.com/webinars/hcl-notes-and-domino-license-cost-reduction-in-the-world-of-dlau/
The introduction of DLAU and the CCB & CCX licensing model caused quite a stir in the HCL community. As a Notes and Domino customer, you may have faced challenges with unexpected user counts and license costs. You probably have questions on how this new licensing approach works and how to benefit from it. Most importantly, you likely have budget constraints and want to save money where possible. Don’t worry, we can help with all of this!
We’ll show you how to fix common misconfigurations that cause higher-than-expected user counts, and how to identify accounts which you can deactivate to save money. There are also frequent patterns that can cause unnecessary cost, like using a person document instead of a mail-in for shared mailboxes. We’ll provide examples and solutions for those as well. And naturally we’ll explain the new licensing model.
Join HCL Ambassador Marc Thomas in this webinar with a special guest appearance from Franz Walder. It will give you the tools and know-how to stay on top of what is going on with Domino licensing. You will be able lower your cost through an optimized configuration and keep it low going forward.
These topics will be covered
- Reducing license cost by finding and fixing misconfigurations and superfluous accounts
- How do CCB and CCX licenses really work?
- Understanding the DLAU tool and how to best utilize it
- Tips for common problem areas, like team mailboxes, functional/test users, etc
- Practical examples and best practices to implement right away
2. overview intro oscillator approach filterbank approach template approach challenges eval summary
introduction
overview
corresponding textbook section
section 9.5
lecture content
• introduction to tempo detection and beat tracking
• overview over basic approaches
• typical challenges
learning objectives
• discuss advantages and disadvantages for different approaches to tempo detection
and beat tracking
• summarize the typical challenges of beat tracking systems
module 9.5: tempo detection 1 / 12
3. overview intro oscillator approach filterbank approach template approach challenges eval summary
introduction
overview
corresponding textbook section
section 9.5
lecture content
• introduction to tempo detection and beat tracking
• overview over basic approaches
• typical challenges
learning objectives
• discuss advantages and disadvantages for different approaches to tempo detection
and beat tracking
• summarize the typical challenges of beat tracking systems
module 9.5: tempo detection 1 / 12
4. overview intro oscillator approach filterbank approach template approach challenges eval summary
tempo detection & beat tracking
problem statement
tempo detection
• detect speed of regular pulse (foot-tapping rate)
beat tracking
• detect the time instances the tempo pulses occur (beat phase)
module 9.5: tempo detection 2 / 12
5. overview intro oscillator approach filterbank approach template approach challenges eval summary
tempo detection & beat tracking
introduction
objectives
1 find the tempo from the
novelty function/onsets
2 find the beat locations
systematic problems:
1 distinguish hierarchical
levels
▶ meter
▶ beat
▶ subbeat/tatum
2 detect beats without onsets
3 recognize onsets without
beats
module 9.5: tempo detection 3 / 12
matlab
source:
plotBeatGrid.m
6. overview intro oscillator approach filterbank approach template approach challenges eval summary
tempo detection & beat tracking
typical beat tracking system
Audio/Midi
Signal
Compute
Novelty Function
or
Onset Times
Compute
Confidence
or
Distance
Update
Current State
Tempo, Beat Phase,
Time
Predict
Next Beat Pos
module 9.5: tempo detection 4 / 12
7. overview intro oscillator approach filterbank approach template approach challenges eval summary
tempo detection & beat tracking
oscillator approach
Beat tracking with an oscillator1
1 initialize pulse generator (tempo estimate, beat position estimate)
2 predict next beat location with pulse
3 adapt acc. to distance (predicted vs. real onset position)
• beat period
• beat phase
4 predict with adapted settings
5 adapt . . .
1
E. W. Large, “Beat Tracking with a Nonlinear Oscillator,” in Proceedings of the 14th International Joint Conference on Artificial Intelligence
(IJCAI), Montreal, Aug. 1995.
module 9.5: tempo detection 5 / 12
8. overview intro oscillator approach filterbank approach template approach challenges eval summary
tempo detection & beat tracking
oscillator approach
Beat tracking with an oscillator1
1 initialize pulse generator (tempo estimate, beat position estimate)
2 predict next beat location with pulse
3 adapt acc. to distance (predicted vs. real onset position)
• beat period
• beat phase
4 predict with adapted settings
5 adapt . . .
1
E. W. Large, “Beat Tracking with a Nonlinear Oscillator,” in Proceedings of the 14th International Joint Conference on Artificial Intelligence
(IJCAI), Montreal, Aug. 1995.
module 9.5: tempo detection 5 / 12
9. overview intro oscillator approach filterbank approach template approach challenges eval summary
tempo detection & beat tracking
oscillator approach
Beat tracking with an oscillator1
1 initialize pulse generator (tempo estimate, beat position estimate)
2 predict next beat location with pulse
3 adapt acc. to distance (predicted vs. real onset position)
• beat period
• beat phase
4 predict with adapted settings
5 adapt . . .
1
E. W. Large, “Beat Tracking with a Nonlinear Oscillator,” in Proceedings of the 14th International Joint Conference on Artificial Intelligence
(IJCAI), Montreal, Aug. 1995.
module 9.5: tempo detection 5 / 12
10. overview intro oscillator approach filterbank approach template approach challenges eval summary
tempo detection & beat tracking
oscillator approach
Beat tracking with an oscillator1
1 initialize pulse generator (tempo estimate, beat position estimate)
2 predict next beat location with pulse
3 adapt acc. to distance (predicted vs. real onset position)
• beat period
• beat phase
4 predict with adapted settings
5 adapt . . .
1
E. W. Large, “Beat Tracking with a Nonlinear Oscillator,” in Proceedings of the 14th International Joint Conference on Artificial Intelligence
(IJCAI), Montreal, Aug. 1995.
module 9.5: tempo detection 5 / 12
11. overview intro oscillator approach filterbank approach template approach challenges eval summary
tempo detection & beat tracking
oscillator approach: initialization
How to estimate the initial tempo
module 9.5: tempo detection 6 / 12
matlab
source:
plotBeatHistogram.m
12. overview intro oscillator approach filterbank approach template approach challenges eval summary
tempo detection & beat tracking
oscillator approach: initialization
How to estimate the initial tempo
location of maximum of ACF of novelty function
maximum of IOI histogram
maximum of beat spectrum/histogram
. . .
module 9.5: tempo detection 6 / 12
matlab
source:
plotBeatHistogram.m
13. overview intro oscillator approach filterbank approach template approach challenges eval summary
tempo detection & beat tracking
multi-agent approach
1 run multiple beat trackers with different parameters
• initial tempo
• initial beat phase
• adaptation speed
2 compute reliability/confidence criteria:
• match beat and onset times
• tempo stability
• majority of different agents
• . . .
3 choose most reliable agent (or path between agents)
module 9.5: tempo detection 7 / 12
14. overview intro oscillator approach filterbank approach template approach challenges eval summary
tempo detection & beat tracking
multi-agent approach
1 run multiple beat trackers with different parameters
• initial tempo
• initial beat phase
• adaptation speed
2 compute reliability/confidence criteria:
• match beat and onset times
• tempo stability
• majority of different agents
• . . .
3 choose most reliable agent (or path between agents)
module 9.5: tempo detection 7 / 12
15. overview intro oscillator approach filterbank approach template approach challenges eval summary
tempo detection & beat tracking
multi-agent approach
1 run multiple beat trackers with different parameters
• initial tempo
• initial beat phase
• adaptation speed
2 compute reliability/confidence criteria:
• match beat and onset times
• tempo stability
• majority of different agents
• . . .
3 choose most reliable agent (or path between agents)
module 9.5: tempo detection 7 / 12
16. overview intro oscillator approach filterbank approach template approach challenges eval summary
tempo detection & beat tracking
multi-agent approach
1 run multiple beat trackers with different parameters
• initial tempo
• initial beat phase
• adaptation speed
2 compute reliability/confidence criteria:
• match beat and onset times
• tempo stability
• majority of different agents
• . . .
3 choose most reliable agent (or path between agents)
module 9.5: tempo detection 7 / 12
17. overview intro oscillator approach filterbank approach template approach challenges eval summary
tempo detection & beat tracking
multi-agent approach
1 run multiple beat trackers with different parameters
• initial tempo
• initial beat phase
• adaptation speed
2 compute reliability/confidence criteria:
• match beat and onset times
• tempo stability
• majority of different agents
• . . .
3 choose most reliable agent (or path between agents)
module 9.5: tempo detection 7 / 12
18. overview intro oscillator approach filterbank approach template approach challenges eval summary
tempo detection & beat tracking
filterbank approach
1 design filterbank (e.g. comb
resonators spaced 1 beat)
2 compute filter output energy
3 pick maximum
plots by Scheirer2
2
E. D. Scheirer, “Tempo and beat analysis of acoustic musical signals,” Journal of the Acoustical Society of America (JASA), vol. 103, no. 1,
pp. 588–601, 1998.
module 9.5: tempo detection 8 / 12
19. overview intro oscillator approach filterbank approach template approach challenges eval summary
tempo detection & beat tracking
filterbank approach
1 design filterbank (e.g. comb
resonators spaced 1 beat)
2 compute filter output energy
3 pick maximum
plots by Scheirer2
2
E. D. Scheirer, “Tempo and beat analysis of acoustic musical signals,” Journal of the Acoustical Society of America (JASA), vol. 103, no. 1,
pp. 588–601, 1998.
module 9.5: tempo detection 8 / 12
20. overview intro oscillator approach filterbank approach template approach challenges eval summary
tempo detection & beat tracking
filterbank approach
1 design filterbank (e.g. comb
resonators spaced 1 beat)
2 compute filter output energy
3 pick maximum
plots by Scheirer2
2
E. D. Scheirer, “Tempo and beat analysis of acoustic musical signals,” Journal of the Acoustical Society of America (JASA), vol. 103, no. 1,
pp. 588–601, 1998.
module 9.5: tempo detection 8 / 12
21. overview intro oscillator approach filterbank approach template approach challenges eval summary
tempo detection & beat tracking
template-based approach
1 define set of template pulses in all tempi
2 compute CCF between novelty function (or its ACF) and all templates
3 choose template with highest correlation as tempo
4 choose lag with highest correlation as beat phase
module 9.5: tempo detection 9 / 12
22. overview intro oscillator approach filterbank approach template approach challenges eval summary
tempo detection & beat tracking
typical problems
1 tempo: detection of double/half tempo (triple, . . . )
2 phase: detection of off-beats
3 tempo & phase: strongly depends on initialization values
4 tempo & phase: only slow adaptation — no sudden tempo changes
example: challenges with adaptation speed
module 9.5: tempo detection 10 / 12
23. overview intro oscillator approach filterbank approach template approach challenges eval summary
tempo detection & beat tracking
evaluation
evaluation of constant tempo
• match within tempo range ⇒ classification metrics
evaluation of beat tracking
• ground truth can be subjective (double/half tempo, deviations)
• each beat matched against ground truth
▶ challenge 1: tolerance window definition (tempo dependent or not?)
▶ challenge 2: slightly different tempo might lead to gap between metrics and
perceptual severity
typical errors
• double/half tempo (sometimes also 3/2 relationships)
• off-beat
• problems with abrupt tempo changes
module 9.5: tempo detection 11 / 12
24. overview intro oscillator approach filterbank approach template approach challenges eval summary
tempo detection & beat tracking
evaluation
evaluation of constant tempo
• match within tempo range ⇒ classification metrics
evaluation of beat tracking
• ground truth can be subjective (double/half tempo, deviations)
• each beat matched against ground truth
▶ challenge 1: tolerance window definition (tempo dependent or not?)
▶ challenge 2: slightly different tempo might lead to gap between metrics and
perceptual severity
typical errors
• double/half tempo (sometimes also 3/2 relationships)
• off-beat
• problems with abrupt tempo changes
module 9.5: tempo detection 11 / 12
25. overview intro oscillator approach filterbank approach template approach challenges eval summary
tempo detection & beat tracking
evaluation
evaluation of constant tempo
• match within tempo range ⇒ classification metrics
evaluation of beat tracking
• ground truth can be subjective (double/half tempo, deviations)
• each beat matched against ground truth
▶ challenge 1: tolerance window definition (tempo dependent or not?)
▶ challenge 2: slightly different tempo might lead to gap between metrics and
perceptual severity
typical errors
• double/half tempo (sometimes also 3/2 relationships)
• off-beat
• problems with abrupt tempo changes
module 9.5: tempo detection 11 / 12
26. overview intro oscillator approach filterbank approach template approach challenges eval summary
summary
lecture content
tempo analysis
• similar to pitch detection on a different scale
▶ periodicity analysis of novelty function
▶ time or spectral domain
typical approaches
• oscillator
• histogram/beat spectrum
• template correlation
main challenges
• double/half tempo
• adaptation to sudden tempo changes
module 9.5: tempo detection 12 / 12