Pitch period spectrogram method for directly displaying pitch period trajectory

Abstract

The invention discloses a pitch cycle spectrogram method for directly displaying pitch cycle tracks. The method first calculates the spectrum and cepstrum by frame, and extracts the pitch period by using the cepstrum. The time corresponding to the cepstrum peak value is the pitch period. Then, the amplitude of the cepstrum is converted into gray scale, and the data points of the cepstrum are converted into pixels. Then, according to the characteristics of similarity, gradient and continuity of the pitch periods of adjacent frames, the pixels of the cepstrum of each frame are drawn side by side along the vertical direction, the horizontal axis is time, and the vertical axis is the cepstrum frequency to form a pitch period spectrum , so that the cepstrum peaks of adjacent frames are close to each other, forming an obvious and regularly changing pitch period trajectory. The invention can directly display the track of the pitch period, not only provides key clues for speech separation, but also can effectively eliminate false cepstrum peaks that deviate from the track of the pitch period, thereby improving the accuracy of extracting the pitch period.

Description

technical field [0001] The invention belongs to the field of speech signal processing, and is mainly used for speech separation, specifically, a pitch cycle spectrogram method for directly displaying pitch cycle traces. Background technique [0002] Speech is the most natural, convenient and common way for human beings to communicate information. Speech recognition has been widely used in human-computer interaction. Although speech recognition can achieve a high recognition rate in a silent environment, in a noisy environment, Speech recognition performance drops significantly. In the actual environment, it is inevitable to be mixed with environmental noise and other people's voice, which will reduce the quality, clarity and intelligibility of the voice signal. The use of voice separation technology can eliminate the interference of other voices and improve the clarity and intelligibility of voice. Speech separation is different from speech denoising. The goal of speech de...

AI Technical Summary

Problems solved by technology

[0023] The purpose of the present invention is to provide a pitch period spectrogram method that directly displays the pitch period trajectory, which is used to solve the problems of low extraction accuracy and poor robustness of the pitch period trajectory in speech separation

Images