Improved pitch period detection algorithm based on main body extension method

Abstract

Due to many outliers in the pitch period detection of noisy speech signal, an improved pitch period detection method is proposed. Firstly, combining spectral subtraction and wavelet transform for noise reduction, we first windowed and framed the useful speech signal sequence obtained after noise reduction for signal endpoint and vowel subject detection. Then, in the second windowing and framing process, the multi window spectrum estimation method was introduced to further reduce the influence of some formants and residual noise on pitch detection. Finally, the improved main body extension method was used for pitch detection. The experimental simulation results show that compared with the traditional pitch detection method, the sample rate of the segment collected by the improved method is 87%, which is at least 5% higher than that of the traditional one. The improved method has higher accuracy of pitch detection.

Keywords

• wavelet transform
• spectral subtraction
• multi-taper spectrum estimation method
• main body extension method
• outliers
• pitch period