Automatic Speech Recognition of Continuous Speech Signal of Gujarati Language Using Machine Learning

In this work we perform automatic recognition of continuous speech signal spoken in Gujarati language using machine learning (ML) technique. For this purpose, from continuous speech signal of sentence we first extract words using short term auto-correlation (STAC) method. Since the signals for each word are large in size, the dimension reduction is done using feature extraction algorithm: mel-frequency discrete wavelet coefficient (MFDWC). Then these features are trained using ML algorithm for recognition of speech.