A comparative study of speaker adaptation methods

For the problem of speaker adaptation in speech recognition, the performance depends on the availability of adaptation data. In this paper, we have compared several existing speaker adaptation methods, viz. maximum likelihood linear regression (MLLR), eigenvoice (EV), eigenspace-based MLLR (EMLLR), segmental eigenvoice (SEV) and hierarchical eigenvoice (HEV) based methods. We also develop a new method by modifying the existing HEV method for achieving further performance improvement in a limited available data scenario. In the sense of availability of adaptation data, the new modified HEV (MHEV) method is shown to perform better than all the existing methods throughout the range of operation except the case of MLLR at the availability of more adaptation data.

Design of speaker identification schemes for large number of speakers (A)

Design of speaker identification schemes for a small number of speakers (around 10) with a high degree of accuracy in controlled environment is a practical proposition today. When the number of speakers is large (say 50–100), many of these schemes cannot be directly extended, as both recognition error and computation time increase monotonically with population size. The feature selection problem is also complex for such schemes. Though there were earlier attempts to rank order features based on statistical distance measures, it has been observed only recently that the best two independent measurements are not the same as the combination in two's for pattern classification. We propose here a systematic approach to the problem using the decision tree or hierarchical classifier with the following objectives: (1) Design of optimal policy at each node of the tree given the tree structure i.e., the tree skeleton and the features to be used at each node. (2) Determination of the optimal feature measurement and decision policy given only the tree skeleton. Applicability of optimization procedures such as dynamic programming in the design of such trees is studied. The experimental results deal with the design of a 50 speaker identification scheme based on this approach.

Voice source characterization using pitch synchronous discrete cosine transform for speaker identification

A characterization of the voice source (VS) signal by the pitch synchronous (PS) discrete cosine transform (DCT) is proposed. With the integrated linear prediction residual (ILPR) as the VS estimate, the PS DCT of the ILPR is evaluated as a feature vector for speaker identification (SID). On TIMIT and YOHO databases, using a Gaussian mixture model (GMM)-based classifier, it performs on par with existing VS-based features. On the NIST 2003 database, fusion with a GMM-based classifier using MFCC features improves the identification accuracy by 12% in absolute terms, proving that the proposed characterization has good promise as a feature for SID studies. (C) 2015 Acoustical Society of America

Speaker verification based on the fusion of speech acoustics and inverted articulatory signals

We propose apractical, feature-level and score-level fusion approach by combining acoustic and estimated articulatory information for both text independent and text dependent speaker verification. From a practical point of view, we study how to improve speaker verification performance by combining dynamic articulatory information with the conventional acoustic features. On text independent speaker verification, we find that concatenating articulatory features obtained from measured speech production data with conventional Mel-frequency cepstral coefficients (MFCCs) improves the performance dramatically. However, since directly measuring articulatory data is not feasible in many real world applications, we also experiment with estimated articulatory features obtained through acoustic-to-articulatory inversion. We explore both feature level and score level fusion methods and find that the overall system performance is significantly enhanced even with estimated articulatory features. Such a performance boost could be due to the inter-speaker variation information embedded in the estimated articulatory features. Since the dynamics of articulation contain important information, we included inverted articulatory trajectories in text dependent speaker verification. We demonstrate that the articulatory constraints introduced by inverted articulatory features help to reject wrong password trials and improve the performance after score level fusion. We evaluate the proposed methods on the X-ray Microbeam database and the RSR 2015 database, respectively, for the aforementioned two tasks. Experimental results show that we achieve more than 15% relative equal error rate reduction for both speaker verification tasks. (C) 2015 Elsevier Ltd. All rights reserved.

A Pattern Recognition Model of Voice-Based Personal Verification Systems for Forensic Applications

Abstract-The success of automatic speaker recognition in laboratory environments suggests applications in forensic science for establishing the Identity of individuals on the basis of features extracted from speech. A theoretical model for such a verification scheme for continuous normaliy distributed featureIss developed. The three cases of using a) single feature, b)multipliendependent measurements of a single feature, and c)multpleindependent features are explored.The number iofndependent features needed for areliable personal identification is computed based on the theoretcal model and an expklatory study of some speech featues.

Joint decoding of multiple speech patterns for robust speech recognition

We are addressing a new problem of improving automatic speech recognition performance, given multiple utterances of patterns from the same class. We have formulated the problem of jointly decoding K multiple patterns given a single Hidden Markov Model. It is shown that such a solution is possible by aligning the K patterns using the proposed Multi Pattern Dynamic Time Warping algorithm followed by the Constrained Multi Pattern Viterbi Algorithm The new formulation is tested in the context of speaker independent isolated word recognition for both clean and noisy patterns. When 10 percent of speech is affected by a burst noise at -5 dB Signal to Noise Ratio (local), it is shown that joint decoding using only two noisy patterns reduces the noisy speech recognition error rate to about 51 percent, when compared to the single pattern decoding using the Viterbi Algorithm. In contrast a simple maximization of individual pattern likelihoods, provides only about 7 percent reduction in error rate.

LZW Based Distance Measures for Spoken Language Identification

We present a new approach to spoken language modeling for language identification (LID) using the Lempel-Ziv-Welch (LZW) algorithm. The LZW technique is applicable to any kind of tokenization of the speech signal. Because of the efficiency of LZW algorithm to obtain variable length symbol strings in the training data, the LZW codebook captures the essentials of a language effectively. We develop two new deterministic measures for LID based on the LZW algorithm namely: (i) Compression ratio score (LZW-CR) and (ii) weighted discriminant score (LZW-WDS). To assess these measures, we consider error-free tokenization of speech as well as artificially induced noise in the tokenization. It is shown that for a 6 language LID task of OGI-TS database with clean tokenization, the new model (LZW-WDS) performs slightly better than the conventional bigram model. For noisy tokenization, which is the more realistic case, LZW-WDS significantly outperforms the bigram technique

Online Estimation of Evolving Human Visual Interest

Regions in video streams attracting human interest contribute significantly to human understanding of the video. Being able to predict salient and informative Regions of Interest (ROIs) through a sequence of eye movements is a challenging problem. Applications such as content-aware retargeting of videos to different aspect ratios while preserving informative regions and smart insertion of dialog (closed-caption text) into the video stream can significantly be improved using the predicted ROIs. We propose an interactive human-in-the-loop framework to model eye movements and predict visual saliency into yet-unseen frames. Eye tracking and video content are used to model visual attention in a manner that accounts for important eye-gaze characteristics such as temporal discontinuities due to sudden eye movements, noise, and behavioral artifacts. A novel statistical-and algorithm-based method gaze buffering is proposed for eye-gaze analysis and its fusion with content-based features. Our robust saliency prediction is instantiated for two challenging and exciting applications. The first application alters video aspect ratios on-the-fly using content-aware video retargeting, thus making them suitable for a variety of display sizes. The second application dynamically localizes active speakers and places dialog captions on-the-fly in the video stream. Our method ensures that dialogs are faithful to active speaker locations and do not interfere with salient content in the video stream. Our framework naturally accommodates personalisation of the application to suit biases and preferences of individual users.

Vocal Tract Length Normalization Features for Audio Search

This paper presents speaker normalization approaches for audio search task. Conventional state-of-the-art feature set, viz., Mel Frequency Cepstral Coefficients (MFCC) is known to contain speaker-specific and linguistic information implicitly. This might create problem for speaker-independent audio search task. In this paper, universal warping-based approach is used for vocal tract length normalization in audio search. In particular, features such as scale transform and warped linear prediction are used to compensate speaker variability in audio matching. The advantage of these features over conventional feature set is that they apply universal frequency warping for both the templates to be matched during audio search. The performance of Scale Transform Cepstral Coefficients (STCC) and Warped Linear Prediction Cepstral Coefficients (WLPCC) are about 3% higher than the state-of-the-art MFCC feature sets on TIMIT database.