Estimating Heart Rate Variation during Walking with SmartphoneUbi NAIST
This document presents a method for estimating heart rate variation during walking using only sensors available on a smartphone. The method constructs a heart rate prediction model using machine learning with inputs of gradient, acceleration amplitude, and estimated oxygen uptake. Evaluation with 18 subjects walking 5 routes showed the method achieved a mean absolute error of less than 7 beats per minute, accurately tracking heart rate variation. Introducing estimated oxygen uptake as a parameter improved accuracy, demonstrating its effectiveness for heart rate prediction.
Estimating Heart Rate Variation during Walking with SmartphoneUbi NAIST
This document presents a method for estimating heart rate variation during walking using only sensors available on a smartphone. The method constructs a heart rate prediction model using machine learning with inputs of gradient, acceleration amplitude, and estimated oxygen uptake. Evaluation with 18 subjects walking 5 routes showed the method achieved a mean absolute error of less than 7 beats per minute, accurately tracking heart rate variation. Introducing estimated oxygen uptake as a parameter improved accuracy, demonstrating its effectiveness for heart rate prediction.
Digital Nature Group at Ars Electronica SummitYoichi Ochiai
The document presents research from the Yoichi Ochiai Laboratory at the University of Tsukuba on their vision of "Digital Nature", which transforms audio-visual media from 2D pixels on flat screens to 3D "pixies" in haptic environments, the production of material existence, the shape of human presence, and human-computer relationships. The ecosystem of Digital Nature will involve interdisciplinary computational projects spanning multimedia systems, graphics, HCI research, fabrication, robotics, art, architecture, materials science, and biology. Examples of artwork and research achievements from their Digital Nature projects are exhibited.
This document discusses and compares several different probabilistic models for sequence labeling tasks, including Hidden Markov Models (HMMs), Maximum Entropy Markov Models (MEMMs), and Conditional Random Fields (CRFs).
It provides mathematical formulations of HMMs, describing how to calculate the most likely label sequence using the Viterbi algorithm. It then introduces MEMMs, which address some limitations of HMMs by incorporating arbitrary, overlapping features. CRFs are presented as an improvement over MEMMs that models the conditional probability of labels given observations, avoiding the label bias problem of MEMMs. The document concludes by describing how to train CRF models using generalized iterative scaling.
Digital Nature Group at Ars Electronica SummitYoichi Ochiai
The document presents research from the Yoichi Ochiai Laboratory at the University of Tsukuba on their vision of "Digital Nature", which transforms audio-visual media from 2D pixels on flat screens to 3D "pixies" in haptic environments, the production of material existence, the shape of human presence, and human-computer relationships. The ecosystem of Digital Nature will involve interdisciplinary computational projects spanning multimedia systems, graphics, HCI research, fabrication, robotics, art, architecture, materials science, and biology. Examples of artwork and research achievements from their Digital Nature projects are exhibited.
This document discusses and compares several different probabilistic models for sequence labeling tasks, including Hidden Markov Models (HMMs), Maximum Entropy Markov Models (MEMMs), and Conditional Random Fields (CRFs).
It provides mathematical formulations of HMMs, describing how to calculate the most likely label sequence using the Viterbi algorithm. It then introduces MEMMs, which address some limitations of HMMs by incorporating arbitrary, overlapping features. CRFs are presented as an improvement over MEMMs that models the conditional probability of labels given observations, avoiding the label bias problem of MEMMs. The document concludes by describing how to train CRF models using generalized iterative scaling.
DeepRemote: A Smart Remote Controller for Intuitive Control through Home Appl...Yuta Takahashi
This document describes DeepRemote, a smart remote controller that uses deep learning for intuitive home appliance selection and control. It consists of a control unit with a camera and buttons and a deep learning unit for appliance recognition. The system was tested for classification accuracy of over 80% on average, response time of under 2 seconds, and faster control times than traditional remotes in user tests. Overall, DeepRemote demonstrates an effective deep learning approach for selecting and controlling home appliances intuitively with a single remote controller.
An Identification Method of IR Signals to Collect Control Logs of Home Applia...Yuta Takahashi
This document proposes a method to identify infrared (IR) signals from home appliances in order to collect control logs. It involves preprocessing raw IR signals into pulse width sequences, comparing signals using mean absolute error and sum absolute error, and constructing statistical models to identify appliance type with 95.5% accuracy and command type with 92% accuracy based on a database of 1,400 signals from 14 appliances. A simple simulation shows identification stability is achieved when the database includes 6 or more signals per appliance. The method could help automatically understand user preferences from appliance usage logs.