A Novel Approach to Speech Signal Segmentation Based on Time-Frequency Analysis

The parameters utilised to establish the starting and end points of relevant pieces in a continuous speech stream directly affect the segmentation accuracy of speech signals. The work's goal is to improve speech/pause segmentation accuracy by frequency-time analysis of speech data. The examination of the values of the mean frequency (in the frequency domain) and short-term energy of the Teager operator function (in the time domain) is used to offer a novel original method for speech/pause segmentation. Due to an additional method to rectify speech/pause segmentation mistakes that was built based on the physiological operation of the respiratory apparatus organs during the production of a continuous speech stream, the suggested solution is distinctive. The performance of the suggested approach has been described, and a brief review of the informative speech signal characteristics used for speech/pause segmentation has been provided. The proposed method has been contrasted with established techniques for speech/pause segmentation for clean and noisy voice signals. The research results have demonstrated that the methods based on the proposed approach produce the best speech/pause segmentation results for both clean and noisy speech signals; the ratio of the short-term energy of the Teager operator function to the mean frequency as an informative parameter ensures maximum relevance to the segmentation problem; and an auxiliary algorithm to correct false states enhances segmentation efficiency.

Keywords: Speech Signal Processing, Speech/Pause Segmentation, Fourier Transform, Teager Energy Operator.