http://www.informatik.uni-ulm.de/ni/staff/FSchwenker/lit/papers/2001-Trentin-NC.pdf

A survey of hybrid ANN/HMM models for automatic speech recognition

The authors describe two systems in which neural network classifiers are merged with dynamic programming (DP) time alignment methods to produce high-performance continuous speech recognizers. One system uses the connectionist Viterbi-training (CVT) procedure, in which a neural network with frame-level outputs is trained using guidance from a time alignment procedure. The other system uses multi-state time-delay neural networks (MS-TDNNs), in which embedded DP time alignment allows network training with only word-level external supervision. The CVT results on the, TI Digits are 99.1% word accuracy and 98.0% string accuracy. The MS-TDNNs are described in detail, with attention focused on their architecture, the training procedure, and results of applying the MS-TDNNs to continuous speaker-dependent alphabet recognition: on two speakers, word accuracy is respectively 97.5% and 89.7%. >

Integrating time alignment and neural networks for high performance continuous speech recognition

An approach to speaker recognition based on feedforward neural models is investigated. Each person known to the system has a personalized neural net that is trained to be active for only that person's speech. By including speech from many people in the training data of each net this approach can directly model differences in people's speech. The chosen architecture and amount of training performed is shown to strongly affect the recognition performance. Large models with two hidden layers are shown to be inferior to models with only a single hidden layer and fewer weights. Recognition performance is shown to be comparable to that of a vector quantization approach based on personalized codebooks. The neural approach outperforms the codebook system for small model sizes but does slightly less well for larger models. A multitransputer implementation used in the training phase is described. Near linear speedup is obtained by splitting the training data to given independent subtasks, and a dynamic allocation scheme is used to assign these tasks to processors. >

Optimisation of neural models for speaker identification

Introduction multisets and fuzzy multisets model logic, rough sets, and fuzzy sets fuzzy cognitive maps - analysis and extensions methods in hard and fuzzy clustering soft-competitive learning paradigms aggregation operations for fusing fuzzy information fuzzy gated neural networks in pattern recognition soft computing technique in kansei (emotional) information processing vagueness in human judgment and decision making chaos and time series analysis a short course for fuzzy set theory.

Soft Computing and Human-Centered Machines

Connected recognition with a recurrent network

Several improvements in the Back-Propagation procedure are proposed to increase training speed, and we discuss their limitations with respect to generalization. performance. The error surface is modeled to avoid local minima and flat areas. The synaptic weights are updated as often as possible. Both the step size and the momentum are dynamically scaled to the largest possible values that do not result in overshooting. Training for the speaker-dependent recognition of the phonemes /b/, /d/ and /g/ has been reduced from 2 days to 1 minnte on an Alliant parallel computer, delivering the same 98.6% recognition performance. With a 55000-connection TDNN, the same algorithm needs 1 hour and 5000 training tokens to recognize the 18 Japanese consonants with 96.7% correct.

Fast back-propagation learning methods for large phonemic neural networks.

The TDNN-LR large-vocabulary and continuous speech recognition system.

This paper is concerned about speaker veri ca tion SV using the sequential probability ratio test SPRT In the SPRT input samples are usually as sumed to be i i d samples from a probability density function because an on line probability computation is required Feature vectors used in speech processing obviously do not satisfy the assumption and there fore the correlation between successive feature vectors has not been considered in conventional SV using the SPRT The correlation can be modeled by the hidden Markov model HMM but unfortunately the HMM can not be directly applied to the SPRT because of statistical dependence of input samples This paper proposes a method of HMM probability computation using the mean eld approximation to resolve this problem where the probability of whole input samples is nominally represented as the product of probability of each sample as if input samples were independent each other

Hidefumi Sawai

Papers

Fast back-propagation learning methods for large phonemic neural networks.

The TDNN-LR large-vocabulary and continuous speech recognition system.

A context-dependent approach for speaker verification using sequential decision.

Time-delay neural network architectures for high-performance speaker-independent recognition.