Short-Time Analysis of Speech Signals

Using MATLAB to perform short-time analysis of speech signals, including voiced/unvoiced classification and pitch period estimation, followed by experimental report writing. The implementation involves MATLAB programming to extract both time-domain and frequency-domain features of speech signals through frame-based processing using windowing functions like Hamming or Hanning windows. For voiced/unvoiced decision, we employ zero-crossing rate analysis and short-time energy calculations - higher zero-crossing rates typically indicate unvoiced sounds while higher energy levels suggest voiced segments. Pitch period estimation is implemented using autocorrelation-based methods or cepstral analysis, where the algorithm detects periodic patterns by finding peaks in the autocorrelation function or cepstrum domain. The experimental report documents the complete analysis methodology, including specific MATLAB functions used (such as xcorr for autocorrelation, fft for spectral analysis), parameter settings (frame size, overlap percentage), obtained results, and conclusions to ensure reproducibility of the research work.