HINDI SPEECH AUDIO VISUAL FEATURE RECOGNITION

  • Sharmila, A.N Mishra, Neeta Awasthy, Vineeta Verma, Sarika Malhotra

Abstract

Automatic Speech Recognition (ASR) system is designed to perform well with some constrained and unfavorable conditions. However, under noisy conditions its performance deteriorates with increase in noise level. Features of Audio-visual of ASR have an important role   in noisy environment. Here, in this research paper, the problem belonging to noisy environment has been dealt by   incorporation of features of Audio-Visual.  Here speech recognition system Hindi is designed with the help of AV features.  The features with discrete wavelet of the lip portion that integrated with the Bark frequency Cepstral coefficients (BFCC) and features of audio have been used for getting better performance of recognition in noisy environments. Pseudo Hue & Intensity of Colour approaches are used for localization lip. Here the classifier is Hidden Markov Model (HMM). Recognition performance with the help of HMM has been found superior to LDA recognizer. For analysis technique of compression of image principal component has been utilized. 

Published
2020-04-07
How to Cite
Sharmila, A.N Mishra, Neeta Awasthy, Vineeta Verma, Sarika Malhotra. (2020). HINDI SPEECH AUDIO VISUAL FEATURE RECOGNITION. International Journal of Advanced Science and Technology, 29(5s), 1734 - 1743. Retrieved from http://sersc.org/journals/index.php/IJAST/article/view/8326