Further results on the recognition of a continuously read natural corpus

Bahl, Lalit R.; Bakis, Raimo; Cohen, Paul S.; Cole, A.; Jelinek, F.; Lewis, B.; Mercer, R.L.

doi:10.1109/icassp.1980.1170862

Cited by 45 publications

(26 citation statements)

References 4 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Third, the acoustic model, implemented by a Deep Neural Network (DNN), transforms the MFCC features into phonemes' probabilities. Context-sensitive phonemes are the norm, triphones [18] being the most common approach. A triphone is a particular phoneme when combined with a particular predecessor and a particular successor.…”

Section: Speech Recognition With Wfstmentioning

confidence: 99%

An ultra low-power hardware accelerator for automatic speech recognition

Yazdani¹,

Segura²,

Arnau³

et al. 2016

2016 49th Annual IEEE/ACM International Symposium on Microarchitecture (MICRO)

View full text Add to dashboard Cite

Abstract-Automatic Speech Recognition (ASR) is becoming increasingly ubiquitous, especially in the mobile segment. Fast and accurate ASR comes at a high energy cost which is not affordable for the tiny power budget of mobile devices. Hardware acceleration can reduce power consumption of ASR systems, while delivering high-performance.In this paper, we present an accelerator for large-vocabulary, speaker-independent, continuous speech recognition. It focuses on the Viterbi search algorithm, that represents the main bottleneck in an ASR system. The proposed design includes innovative techniques to improve the memory subsystem, since memory is identified as the main bottleneck for performance and power in the design of these accelerators. We propose a prefetching scheme tailored to the needs of an ASR system that hides main memory latency for a large fraction of the memory accesses with a negligible impact on area. In addition, we introduce a novel bandwidth saving technique that removes 20% of the off-chip memory accesses issued during the Viterbi search.The proposed design outperforms software implementations running on the CPU by orders of magnitude and achieves 1.7x speedup over a highly optimized CUDA implementation running on a high-end Geforce GTX 980 GPU, while reducing by two orders of magnitude (287x) the energy required to convert the speech into text.

show abstract

Section: Speech Recognition With Wfstmentioning

confidence: 99%

An ultra low-power hardware accelerator for automatic speech recognition

Yazdani¹,

Segura²,

Arnau³

et al. 2016

2016 49th Annual IEEE/ACM International Symposium on Microarchitecture (MICRO)

View full text Add to dashboard Cite

show abstract

“…Some of the possibilities include sub-phoneme units, phones with right or left context, biphones, diphones [2] and variations [3], dyads or transemes [4], avents [5], triphones [6],demisyllables [7], whole words and phrases. Current research in psychoacoustics and psycholinguistics suggest that the syllable might be a basic unit of human speech perception.…”

Section: Introductionmentioning

confidence: 99%

A Review on Acoustic Phonetic Approach for Marathi Speech Recognition

Pawar¹

2012

IJCA

View full text Add to dashboard Cite

This paper discusses the phoneme used in Marathi language as a possible basic unit of speech recognition, for which there is some empirical psychoacoustic support in the case of human and some engineering justification in the case of machines striving to imitate human abilities. For the purpose of the research described in this paper, a basic unit of speech recognition is the intermediate form of speech information around which much of the recognition processing is organized for human beings or for machines. The general opinion of phonetician and psycholinguists is that there is indeed such a unit with relatively few distinct types 1 . For this research a basic unit is ideally an output of acoustic-phonetic processing and an input to the lexical processing stages.

show abstract

“…Both methods are based on the computation of emission probabilities directly on the training data during the forward-backward training algorithm. The first method characterizes the difference between two distributions by means of a divergence criteria derived from the one proposed in [4]: (1) where is the kth training occurrence of model , is the isolated emission probability of by model , and is the number of training occurrences of model . The underlying idea of our method is that a model is considered to be well trained if there is a sufficient number of examples for this model.…”

Section: The Clustering Techniquesmentioning

confidence: 99%

“…(1). To guide the clustering and to avoid having too few/too many clusters per class, the final number of clusters per class is constrained between two thresholds: Nmin and Nmax.…”

Section: Context Modeling and Clustering In Continuous Speech Recognimentioning

confidence: 99%

See 1 more Smart Citation

Context modeling and clustering in continuous speech recognition

Junqua

Vassallo²

Proceeding of Fourth International Conference on Spoken Language Processing. ICSLP '96

View full text Add to dashboard Cite

In this paper, we report on the performance of two variants of wellknown statistical-based clustering techniques and present an evaluation on the TIMIT and TI-Digit databases. A clustering approach which 1) is based on a divergence criterion, 2) separates "good" and "bad" models using a class-dependent adjustable threshold on the number of examples per model, and 3) guides the clustering by limiting the number of models per class between two constants Nmin and Nmax, gave the best results.On the TI-Digit database, the combination of triphone modeling and divergence-based clustering yielded greater accuracy than that obtained with word models for a similar system complexity.

show abstract

Further results on the recognition of a continuously read natural corpus

Cited by 45 publications

References 4 publications

An ultra low-power hardware accelerator for automatic speech recognition

An ultra low-power hardware accelerator for automatic speech recognition

A Review on Acoustic Phonetic Approach for Marathi Speech Recognition

Context modeling and clustering in continuous speech recognition

Contact Info

Product

Resources

About