Experiments in hearing

A new type of End-to-End system for text-dependent speaker verification is presented in this paper. Previously, using the phonetically discriminative/speaker discriminative DNNs as feature extractors for speaker verification has shown promising results. The extracted frame-level (DNN bottleneck, posterior or d-vector) features are equally weighted and aggregated to compute an utterance-level speaker representation (d-vector or i-vector). In this work we use speaker discriminative CNNs to extract the noise-robust frame-level features. These features are smartly combined to form an utterance-level speaker vector through an attention mechanism. The proposed attention model takes the speaker discriminative information and the phonetic information to learn the weights. The whole system, including the CNN and attention model, is joint optimized using an end-to-end criterion. The training algorithm imitates exactly the evaluation process --- directly mapping a test utterance and a few target speaker utterances into a single verification score. The algorithm can automatically select the most similar impostor for each target speaker to train the network. We demonstrated the effectiveness of the proposed end-to-end system on Windows $10$ "Hey Cortana" speaker verification task.

End-to-End Attention based Text-Dependent Speaker Verification

A new type of End-to-End system for text-dependent speaker verification is presented in this paper. Previously, using the phonetic discriminate/speaker discriminate DNN as a feature extractor for speaker verification has shown promising results. The extracted frame-level (bottleneck, posterior or d-vector) features are equally weighted and aggregated to compute an utterance-level speaker representation (d-vector or i-vector). In this work we use a speaker discriminate CNN to extract the noise-robust frame-level features. These features are smartly combined to form an utterance-level speaker vector through an attention mechanism. The proposed attention model takes the speaker discriminate information and the phonetic information to learn the weights. The whole system, including the CNN and attention model, is joint optimized using an end-to-end criterion. The training algorithm imitates exactly the evaluation process — directly mapping a test utterance and a few target speaker utterances into a single verification score. The algorithm can smartly select the most similar impostor for each target speaker to train the network. We demonstrated the effectiveness of the proposed end-to-end system on Windows 10 “Hey Cortana” speaker verification task.

/pdf/end-to-end-attention-based-text-dependent-speaker-3vgx6ss7ze.pdf

End-to-End attention based text-dependent speaker verification

Sleep stages classification based on heart rate variability and random forest

http://cs.joensuu.fi/pages/tkinnu/webpage/pdf/multitaper_speechcomm.pdf

Multitaper MFCC and PLP features for speaker verification using i-vectors

This paper investigates several feature normalization techniques for use in an i-vector speaker verification system based on a mixture probabilistic linear discriminant analysis (PLDA) model. The objective of the feature normalization technique is to compensate for the effects of environmental mismatch. Here, we study short-time Gaussianization (STG), short-time mean and variance normalization (STMVN), and short-time mean and scale normalization (STMSN) techniques. Our goal is to compare the performance of the above mentioned feature normalization techniques on the telephone (det5) and microphone speech (det1, det2, det3 and det4) of the NIST SRE 2010 corpora. Experimental results show that the performances of the STMVN and STMSN techniques are comparable to that of the STG technique.

Comparative evaluation of feature normalization techniques for speaker verification

Robust feature extraction based on an asymmetric level-dependent auditory filterbank and a subband spectrum enhancement technique

Deep Neural Network based Text-Dependent Speaker Verification : Preliminary Results.

In this paper we study low-variance multi-taper spectrum estimation methods to compute the mel-frequency cepstral coefficient (MFCC) features for robust speech recognition. In speech recognition, MFCC features are usually computed from a Hamming-windowed DFT spectrum. Although windowing helps in reducing the bias of the spectrum, but variance remains high. Multitaper spectrum estimation methods can be used to correct the shortcomings of single taper (or window) spectrum estimation methods. Experimental results on the AURORA-2 corpus show that the multi-taper methods, specifically the multi-peak multi-taper method, perform better compared to the Hamming-windowed spectrum estimation method.

A study of low-variance multi-taper features for distributed speech recognition

This paper presents regularized minimum variance distortion-less response (MVDR)-based cepstral features for robust continuous speech recognition. The mel-frequency cepstral coefficient (MFCC) features, widely used in speech recognition tasks, are usually computed from a direct spectrum estimate, that is, the squared magnitude of the discrete Fourier transform (DFT) of speech frames. Direct spectrum estimation methods (also known as nonparametric estimators) perform poorly under noisy and adverse conditions. To reduce this performance drop we propose to increase robustness of the speech recognition system by extracting more robust features based on the regularized MVDR technique. The proposed method, when evaluated on the AURORA-4 speech recognition task, provides an average relative improvement in word accuracy of 11.3%, 6.1%, and 5.2% over the conventional MFCC, PLP, MVDR and PMVDR-based MFCC features, respectively.

Jahangir Alam

Papers

Comparative evaluation of feature normalization techniques for speaker verification

Robust feature extraction based on an asymmetric level-dependent auditory filterbank and a subband spectrum enhancement technique

Deep Neural Network based Text-Dependent Speaker Verification : Preliminary Results.

A study of low-variance multi-taper features for distributed speech recognition

Speech recognition using regularized minimum variance distortionless response spectrum estimation-based cepstral features