6. 1. CASA-Based Robust Speaker Identification
(Computational Auditory Scene Analysis)
7. 2. Independent component analysis and MLLR
transformation for speaker identification
• Independent Component Analysis (ICA).
• Principle Component Analysis (PCA).
3. Towards noise –robust speaker recognition
using probabilistic linear discriminant analysis
• Probabilistic linear discriminant analysis
• Additive noise
8. 4. Weighted LDA techniques for i-vector
based speaker verification.
• Improving i-vector speaker verification in presence of
high inter session variability.
• Interview-interview condition.
• Telephone-telephone condition.
5. An Overview of Speaker Identification:
Accuracy and Robustness Issues
Two methods
• Speaker identification
• Speaker verification
9. 6. Cross-pollination of normalization techniques
from speaker to face authentication using
Gaussian mixture models.
10. 7.Front-End Factor Analysis for Speaker
Verification
• This paper proposed new way of combining JFA and
SVM’s for speaker verification.
8. Parallel transformation network feature for
speaker recognition
• TN features with SVM modeling-method in order to
become language independent and overcome the need
for accurate speech recognition.
11. 9. Statistical Pattern Recognition Techniques for
Speaker Verification
10. Speaker Identification within Whispered Speech
Audio Streams
Whisper is an alternative speech production mode used by
subjects in natural conversation to protect the privacy.
Whispered speech is a natural mode of speech information.
12. 11. A comparison of approaches for modeling prosodic
features in speaker recognition.
• It address the task of text-independent speaker
verification.
• Prosodic features.
12.Fusion Methods for Boosting Performance of
Speaker Identification Systems
1. feature extraction.
2.classification tasks.
13. 13. Source-normalized LDA for robust speaker
recognition using i-vectors from multiple speech
sources
• Improves the robustness of i-vector-based speaker
recognition.
• An source-normalized algorithm to improves
robustness of i-vector-based-speaker recognition.
14. A study on Universal Background Model training
in Speaker Verification
• Systematic analyze of speaker verification system
performance.
• Rigorous methods like IFS scheme is used to
estimate similarity.
14. 15. Speaker Identification Using Instantaneous
Frequencies
• Introduction of new set of descriptors that capture the
identity of speaker well.
• Provides robustness with respect to changes in
recording channel and speaking style.
16. Codebook Design Method for Noise Robust
Speaker Identification based on Genetic
Algorithm
• To designing a codebook for noise robust speaker,
Genetic algorithm is proposed.
16. 17. Enhanced speaker recognition based on intra-
modal fusion and accent modeling.
• Intra-modal fusion.
• Accent modeling.
18. Discriminant NAP for SVM Speaker Recognition
• Nuisance Attribute Projection (NAP) provides an
effective method of removing the unwanted session
variability in a Support Vector Machine (SVM) based
speaker recognition system by removing the principal
components of this variability.
17. 19. A Speech-and-Speaker Identification System:
Feature Extraction, Description and Classification of
Speech-Signal Image
• A speech-and-speaker (SAS) identification system
based on spoken Arabic digit recognition.
18. 20. In-Set/Out-of-Set Speaker Recognition Under
Sparse Enrollment
• The problem of in-set speaker recognition is
addressed with the constraints of low enrollment (5 s)
and test material (2–8 s) and in-set group sizes
ranging from 15–45 speakers.
• An algorithm is proposed that uses an in-set
speaker’s cohort set to make up for the sparse (e.g., 5
s per speaker) enrollment data.
21. Analysis of Speech Recognition Techniques for
use in a Non-Speech Sound Recognition System
• Analysis the different techniques used for speech
recognition and identifies those that can be used for
non-speech sound recognition
19. 22. Speaker verification for home security system
• A reliable speaker verification algorithm is used in
home security.
20. 23. An Efficient Scoring Algorithm for Gaussian
Mixture Model Based Speaker Identification
• The use of GMM for speaker identification was
shown to provide superior performance
Graphical illustration of the observation vector recording
21. 24. Speaker Recognition: A Tutorial
• Speech processing is a diverse field with many
applications.
22. 25. Speaker Identification Based on the Use of
Robust Cepstral Features Obtained from Pole-Zero
Transfer Functions
• An attempt made to alleviate mismatch in the training and
testing conditions.
• Proposed a new feature called linear predictive ceptrum
derived by pole-zero function.
26. Speaker Verification Using Mixture Decomposition
Discrimination
• Mixture decomposition discrimination (MDD) is based on
the idea that, when modeling speech using hidden Markov
models (HMM), different speakers speaking the same word
would cause different HMM mixture components to
dominate.
24. 28. Unsupervised Speaker Recognition Based on
Competition Between Self-Organizing Maps
• Clustering the speaker from unlabeled and unsegmented
conversation, when no priori knowledge about the identity
of the participants is given.
29. Speaker Recognition with Polynomial Classifiers
• Polynomial –based classifier to achieve high accuracy at low
complexity.
- It has several advantages.
1. Polynomial classifier scoring yields a system which is highly
computationally scalable with the number of speakers.
2. A new training algorithm is proposed which is discriminative,
handles large data sets, and has low memory usage.
3. The output of the polynomial classifier is easily incorporated
into a statistical framework allowing it to be combined with
other techniques such as HMM.
25. 30. Automatic Verbal Information Verification for
User Authentication
An example of verbal information verification by asking
sequential questions.
26. Issues and challenges…
• Robustness
• Portability
• Adaptation
• Language modeling
• Confidence measure
• Out of vocabulary words
• Prosody
27. Conclusion…
• Problems are still with speaker-generated variability and
variability in channel and recording conditions.
• It is very important to investigate feature parameters that
are stable over time, insensitive to the variation of
speaking manner, including the speaking rate and level,
and robust against variations in voice quality due to
causes such as voice disguise or colds.
• Studies on ways to automatically extract the speech
periods of each person separately from a dialogue
involving more than two people have recently appeared as
an extension of speaker recognition technology.