Isolated speech recognition using Fuzzy C Means technique

Automatic speech recognition is one of the challenging area in the field of speech signal processing. Automatic speech recognition technology converts speech signal into text. This paper presents the implementation of isolated kannada word recognizer using Vector Quantization (VQ) and Fuzzy-C Means (FCM) techniques. The paper compares and contrasts the recognition accuracies of FCM and k-means techniques. It also highlights the importance of the fuzziness parameter `m' in the FCM technique for a range of `m' (fuzzifier) values between 1.5 to 2. The simulation analysis shows the performance of the FCM is dependent on Fuzzy parameter (`m'). It is observed that the recognition accuracies are better for FCM than VQ for clean and noisy speech signals. The results are tested and evaluated for both speaker dependent and independent speech signals. It is clear from the simulation that the recognition accuracies are better for the values of `m' between 1.8 to 2.0. It also highlights the mathematical reason for the better accuracy of speech recognition process. Recognition accuracy obtained is 90 to 95% for clean and speaker dependent signals. 20 to 25% of recognition accuracy is obtained for noisy and speaker independent signals.