Speech recognition using variable-length frame overlaps by intelligent fuzzy control

Abstract

This work presents a fuzzy logic control (FLC) mechanism to determine a variable-length frame overlap between two consecutive frames for speech recognition. Compared with the conventional fixed-length frame overlapping method, the proposed mechanism can acquire more accurate feature parameters of speech samples when performing the procedure of feature extraction by taking the LPC residual error into consideration, and thus ensures the robustness of speech recognition. The proposed mechanism is conceptually simple and effective and computationally inexpensive. Experiment results indicate that the recognition performance of the proposed variable-length frame overlapping mechanism under FLC regulation is apparently superior to that of the conventional approach with each frame overlap set fixed.