基于DTW的音乐信号特征识别系统研究

Based on the background of feature extraction and recognition of music signal, a music signal feature recognition system based on DTW is proposed. This system carry onweighted processing onthe signalwith the method of signal pre-weighting, signal framing, and the Hamming window function. Double threshold endpoint detection is realized by detecting short-term energy and short-term zero crossing rate. Take the logarithm of the power spectrum obtained by Fourier transform through the filter bank to obtain the MFCC coefficient of each frame. DTW algorithm is used to realize the recognition and training mode of the music recognition system. The simulation results show that the music recognition system proposed can effectively eliminate the noise in the signal, accurately detect the end points of the speech fragments, music recognition and reproduction.