| Music,as a kind of mental expression of people,begin to be widely prevalent in people’s daily lives.As early as the Eastern Han Dynasty,Xu Shen pointed out the relationship between sound and music in the Chinese dictionary " Shuo Wen Jie Zi" which is the first dictionary that systematically analysis the Chinese font and origin in China according to whether it is necessary to create by human’s mental imagination:"Music is a kind of sound,originated from our mind and expressed through the outside,that is the true music." Music is influenced by its own characteristics and has a strong correlation with mathematics and physics.With the increasingly powerful computing capability of computer itself,and the development cycle of related hardware and software gradually be shortened,modern music has a close relationship with the computer field.This article focuses on the detection of music rhythm.In the extraction of the previous basic features,take the problem into consideration that music signal changes faster than the simple speech signal.Therefore,the second-order differential zero-crossing rate and the thirdorder differential zero-crossing rate are proposed to reflect the speed of a song.On the basis of obtaining simple features,in the subsequent part of complex features estimation,this thesis discusses the rhythm from two aspects,one is the division of music beat,and the other is the estimation of the speed of music.In this thesis,the music data is analyzed from the perspective of time domain and frequency domain respectively.After being given the transformation ways between time domain,frequency domain,cepstral domain and power spectrum,experiments are carried out to obtain a variety of features of music aiming at the current mainstream method of extracting the basic features of audio signals.Subsequently,with the help of source separation method that is combined with the CQT detection method,the relative position of the beat point is obtained through the period estimation of the music signal.Through the construction of the implicit Markov model and the experiment of the algorithm,proposed a creative use of HMM for detecting and identifying fast music,slow music and medium speed music.Finally,combined with the practical application of life requirements,using the extracted features as a basis to achieve the control of dance robot.This thesis has completed the extraction of the basic music features from the waveform file,and has made some progress in the research on the complex rhythm features.The thesis also gives some research trend suggestions for the further study in the direction of research content and the improvement of the algorithm in future. |
