Research on Intelligent Vocal Music Training System Based on Wavelet Transform

Abstract

In the process of vocal music learning, most of the learning is through the traditional learning mode, that is, learning knowledge in class. For vocal music lovers who want to learn vocal music knowledge, vocal music learning has many requirements on the arrangement of learning place, time and personnel. The purpose of this paper is to develop and design an intelligent training system for vocal music. The flow of audio fingerprint extraction algorithm based on wavelet transform is to transform audio signals into spectrograms. Then, according to the characteristics that wavelet transform can better represent picture information in both time domain and frequency domain, wavelet transform is used for a series of processing of spectrograms. Finally, a more compact fingerprint is obtained by dimension reduction algorithm. This paper applies this algorithm to vocal intelligent training system for the first time. When the humming quality is high, the detection rate can reach more than 90%.