Speech Pitch Detection Based On Multi-wavelet And The Research On Objective Evaluation Of Songs

The Pitch detection is important in speech signal procession, for instance, speech compression coding, speech recognition and speech synthesis, etc. The accurateness and reliability of pitch period will influence the performance of the whole system seriously. But it is difficult in detecting pitch accurately because many factors affect the detection.This paper aims at looking for a kind of robust pitch detection algorithm. Based on the basic theory of the digital speech signal process, the methods of speech signal process are expounded. Secondly, a more systematic analysis and comparison among some typical methods of pitch detection is made. Thirdly, this paper describes detailedly the basic knowledge of the invariant sets Multi-wavelet, presents the calculation of filter, the algorithm of decomposing and reconstruction. Finally, considering the invariant sets Multi-wavelet perfectly combine smoothness, short support, orthogonality, symmetry, etc. we combine the invariant sets Multi-wavelet transform and autocorrelation; propose a kind of new pitch detection algorithm. The main idea is: Firstly, we decompose the original speech signal with the invariant sets Multi-wavelet, get the low-frequency band signals with abundant pitch information and stronger periodicity. Secondly, base on the low-frequency band signals, we adopt the normal autocorrelation method to detect pitch period.Finish designing program with Matlab on the computer and carry on the simulation experiment. The results show that the method of pitch detection is superior to the traditional Autocorrelation Function and the wavelet transform in measurement accuracy, robustness to noise; it could adapt to the pitch detection in the low SNR environments.Finally, the paper has a try to the evaluation method of songs based on the BP ANN. The evaluation of speech is being studied and developed; at current, there are no complete theories and methods. This paper put the emphasis one the research of the objective evaluation of songs. Using analytical technology of speech signal, the author has developed a computer objective evaluation method to evaluate the overall effectiveness of songs. The experiment's result show that the objective evaluations accord with the subjective evaluations in the main.