Feature Extraction and Classification of Speech Signal Using Hidden Markov-Gaussian Mixture Model (HM-GMM) for Driving the Rehabilitative Aids

For the last few decade, an increase in experimental research has been carried-out to help the people with different-level of disabilities, who seek help towards the mobility or movement or transport aids. Hence, in our study, we have developed a system which has the ability to classify the human speech signal into corresponding commands for triggering a rehabilitative aid such as wheelchair for helping the people seek for a movement from one place to other. For this purpose, we have introduced a classification model which uses hidden Markov-Gaussian mixture, for classifying a speech signal especially the words for driving a wheelchair. To improve the efficiency of the proposed classification model, the classification was performed based on the features extracted from the speech signal. Ten volunteers were used in this study to validate the result and it is found to have an accuracy of about 98% to classify the word into their corresponding command. Hence, we propose this model as a good tool for helping the people with differently-disabled people.