Speech to Text Conveter in Python

A real time speech to text conversion system converts the spoken words into text form exactly in the similar way that the user pronounces. We created a real time speech recognition system that was tested in real time noiseous environment. We used the design of a bidirectional nonstationary Kalman filter to enhance the ability of this Real time speech recognition system. Bidirectional Kalman filter has been proved to be the best noise estimator in nonstationary noiseous environment. Real time speech to text conversion system introduces conversion of the uttered words instantly after the utterance. The purpose of this project was to introduce a new speech recognition system that is computationally simple and more robust to noise than the HMM based speech recognition system. We have used our own created database for its flexibility and TIDIGIT database for its accuracy comparison with the HMM based speech recognition system. MFCC features of speech sample were calculated and words were distinguished according to the feature matching of each sampled word. System was tested in different noise conditions and we obtained overall word accuracy of 90%.