Comparison of conventional methods and deep belief networks for isolated word recognition

Author

Pradeep, R. ; Kumaraswamy, R.

Author_Institution

Dept. of Electron. & Commun., Siddaganga Inst. of Technol., Tumkur, India

fYear

2014

fDate

10-12 Oct. 2014

Firstpage

1

Lastpage

5

Abstract

A comparative analysis of the use of conventional methods and Deep Belief Networks (DBN) for speaker independent Isolated Word Recognition on small vocabulary is discussed in this paper. The conventional methods of speech recognition include HMM/GMM framework and Multilayer Perceptrons (MLPs). Features from the speech frames are used to train MLPs using back-propagation. The features that are extracted are 12^th order LPCs and 39 dimensional MFCCs for each frame. The stacked Restricted Boltzmann Machines (RBM) constitute a Deep Belief Networks (DBNs). The DBN learning procedure undergoes a pre-training stage and a fine-tuning stage. DBNs gave a higher performance as compared with the conventional methods with an accuracy of approximately 93% for Isolated Word Recognition using MFCC features.

Keywords

Boltzmann machines; Gaussian processes; backpropagation; feature extraction; hidden Markov models; multilayer perceptrons; speech recognition; DBN learning procedure; GMM framework; Gaussian mixture model; HMM framework; MFCC features; MLP; back-propagation; deep belief networks; feature extraction; fine-tuning stage; hidden Markov models; multilayer perceptrons; pre-training stage; restricted Boltzmann machines; speaker independent isolated word recognition; speech recognition; Equations; Feeds; Hidden Markov models; Mathematical model; Neural networks; Speech; Speech recognition; Deep Belief Networks; Isolated Word Recognition; Multilayer Perceptrons; Restricted Boltzmann Machines;

fLanguage

English

Publisher

ieee

Conference_Titel

Communication, Signal Processing and Networking (NCCSN), 2014 National Conference on

Conference_Location

Palakkad

Type

conf

DOI

10.1109/NCCSN.2014.7001147

Filename

7001147