Speech recognition using neural networks

Khan, Shahbaz; Sharma, Gaurav; Rao, P. Rama Koteswara

doi:10.1109/icit.2000.854165

Cited by 3 publications

(3 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A Volterra analysis of the second MLP in this case would reveal the articulatory-temporal patterns that are learned for each of the phonemes. Similar hierarchical systems have been previously studied where the second classifier is a RNN [33] or a CRF model [18].…”

Section: A Choice Of the Subword Unitsmentioning

confidence: 99%

“…In a prior work [33], the hierarchical estimation of the phoneme posterior probabilities using an RNN was investigated. The first stage of the hierarchical system consists of an MLP trained using the power spectrum of the speech.…”

Section: G4: Hierarchy Using Rnnmentioning

confidence: 99%

“…The linear activation values at the bottleneck layer are used as features in Tandem based speech [31] T nil -HMM G4 [33] any 1:t P RNN G5 [18] [19] any T P CRF G6 [12] [20] any 230 ms P MLP recognition. The processing from the input to the compression layer can be likened to the first MLP in the hierarchical system, and the processing from the compression layer to the output layer can be likened to the second MLP.…”

Section: G7: Bottleneck Featuresmentioning

confidence: 99%

See 2 more Smart Citations

Analysis of MLP-Based Hierarchical Phoneme Posterior Probability Estimator

Pinto

Garimella

Magimai-Doss

et al. 2011

IEEE Trans. Audio Speech Lang. Process.

View full text Add to dashboard Cite

Abstract-We analyze a simple hierarchical architecture consisting of two multilayer perceptron (MLP) classifiers in tandem to estimate the phonetic class conditional probabilities. In this hierarchical setup, the first MLP classifier is trained using standard acoustic features. The second MLP is trained using the posterior probabilities of phonemes estimated by the first, but with a long temporal context of around 150-230 ms. Through extensive phoneme recognition experiments, and the analysis of the trained second MLP using Volterra series, we show that (a) the hierarchical system yields higher phoneme recognition accuracies -an absolute improvement of 3.5% and 9.3% on TIMIT and CTS respectively -over the conventional single MLP based system, (b) there exists useful information in the temporal trajectories of the posterior feature space, spanning around 230 ms of context, (c) the second MLP learns the phonetic temporal patterns in the posterior features, which include the phonetic confusions at the output of the first MLP as well as the phonotactics of the language as observed in the training data, and (d) the second MLP classifier requires fewer number of parameters and can be trained using lesser amount of training data.

show abstract

Section: A Choice Of the Subword Unitsmentioning

confidence: 99%

Section: G4: Hierarchy Using Rnnmentioning

confidence: 99%

Section: G7: Bottleneck Featuresmentioning

confidence: 99%

See 1 more Smart Citation

Analysis of MLP-Based Hierarchical Phoneme Posterior Probability Estimator

Pinto

Garimella

Magimai-Doss

et al. 2011

IEEE Trans. Audio Speech Lang. Process.

View full text Add to dashboard Cite

show abstract

Speech recognition of deaf and hard of hearing people using hybrid neural network

Jeyalakshmi

Krishnamurthi²,

Revathi

2010

2010 2nd International Conference on Mechanical and Electronics Engineering

View full text Add to dashboard Cite

This paper describes isolated word recognition of deaf students by unsupervised and supervised neural network.Compared to normal speech, there is high variability in deaf speech and by hearing once we couldn't understand it. By the use of proposed method deaf people can make use of all voice operated devices. In this paper we use combination of SOFM and BPN neural network for recognition. Initially the input is sampled, filtered, windowed and Perceptual Linear Predictive Coefficients are determined for each frame. These coefficients are applied as input to the SOFM neural network. The output of this network is given to BPN neural network comprising of 3 layers for learning. The network has been trained with five words uttered by five different deaf persons in the age group of 5-10 years. Another set of same five words uttered by same five deaf persons were used for test purposes. The recognition results for the word one, three, four is 50 to 60% and for five is 50% ... But the recognition results for word two are onlyl0% since the variability is high for two.The results can be improved by varying the parameters of the hybrid neural network.

show abstract

Multilayer Perceptron Based Hierarchical Acoustic Modeling for Automatic Speech Recognition

Pinto¹

2010

View full text Add to dashboard Cite

Speech recognition using neural networks

Cited by 3 publications

References 21 publications

Analysis of MLP-Based Hierarchical Phoneme Posterior Probability Estimator

Analysis of MLP-Based Hierarchical Phoneme Posterior Probability Estimator

Speech recognition of deaf and hard of hearing people using hybrid neural network

Multilayer Perceptron Based Hierarchical Acoustic Modeling for Automatic Speech Recognition

Contact Info

Product

Resources

About