Isolated word speech recognition using larynx vibrations

Loss of speech is largely found in people due to various reasons such as trauma, accidents, paralysis, or post medical treatment. In this era of technological innovation, it is absolutely essential to develop certain low cost techniques to help the speech impaired patients. Present research proposes novel methods for analyzing Speech using Mel Frequency Cepstral Coefficient (MFCC) Vectors and Machine Learning techniques like Hidden Markov Model (HMM) for speech recognition. The developed scheme, proposes the use and modification of such techniques for using Vocal Vibrations captured through transducers to recognize speech of the speech impaired patients with a mean accuracy of 85%. This study holds promising grounds for the speech impaired patients (mis-articulate patients) and future mic-less communication in highly noisy environments.