针对波形音乐文件信号复杂多变的特点, 提出一种适用波形音乐文件的特征提取方法。首先, 在时域上对音乐信号进行去噪, 通过高斯低通滤波器过滤高频信号, 获取信号的包络线, 通过快速傅里叶变换在频域上进行峰值检测, 提取出音符及其特征。然后利用加窗移动匹配算法提取小节及其特征, 利用相邻小节间的相似性划分乐段, 获得具有情感因素的乐段特征。将乐段特征输入到RAN情感识别器进行情感识别, 最终确定整首音乐的情感特征。实验结果表明, 该方法可以对不同类型的音乐进行特征提取。
This paper proposes a feature extraction method for waveform music files based on their complex and changeable signals.First, we denoise the music signals in the time domain.The signal envelope is got by the Gaussian low-pass filter for the high frequency signals.And we use FFT algorithm in frequency domain on the peak detection to extract the notes and the feature.Based on the strength laws of the notes in the bars, we extract bars and bars eigenvector by using the window moving match algorithm, and then mark off the period according to the similarity between bars, and get period feature with emotional factors.The period feature is input into RAN neural network emotion recognizer for emotion recognition of certain periods, and to decide the emotion feature of the whole music.The experimental results show that this method is effective in extracting features of different types of music.
[1]Cope D.Recombinant music:using the computer to explore musical style[J].Computer, 1991, 24(7):22-28.
[2]Nagashima T, Kawashima J.Experimental study on arranging music by chaotic neural network[J].International Journal of Intelligent Systems, 1997, 12(4):323-339.
[3]Johnson M L.Toward an expert system for expressive musical performance[J].Computer, 1991, 24(7):30-34.
[4]Kostek B.Computer-based recognition of musical phrases using the rough-set approach[J].Information Sciences, 1998, 104(1):15-30.
[5]Katayose H, Fukuoka T, Takami K, et al.Expression extraction in virtuoso music performances[C].Proceedings of the 10th International Conference on Pattern Recognition, 1990:1:780-784.
[6]Wang Q, Saiwaki N, Nishida S.An approach to automatic motion synthesis for 3D spring model with music factors[C].Proceedings of IEEE International Conference on System, Man, and Cybernetics, 1999:4:224-229.
[7]Yanase T, Takasu A, Adachi J.Phrase based feature extraction for musical information retrieval[C].Proceedings of 1999 IEEE Pacific Rim Conference on Communications, Computers and Signal Processing, 1999:396-399.
[8]Kashino K, Tanaka H.Asound source separation system with the ability of automatic tone modeling[C].Proceedings of the International Computer Music Accociation Conference on International, 1993:248-255.
[9]Alghoniemy M, Tewfik A H.Rhythm and periodicity detection in polyphonic music[C].Proceedings of 1999 IEEE 3th Workshop on Multimedia Signal Processing, 1999:185-190.
[10]Scheirer E D.Tempo and beat analysis of acoustic musical signals[J].The Journal of the Acoustical Society of America, 1998, 103(1):588-601.
[11]Whiteley N, Cemgil A T, Godsill S.Bayesian modelling of temporal structure in musical audio[C].Proceedings of the 7th International Conference on the Music Information Retrieval, 2006:29-34.
[12]Whiteley N, Cemgil A T, Godsill S.Sequential inference of rhythmic structure in musical audio[C].Proceedings of IEEE International Conference on Acoustics, Speech and Signal Processing, 2007:1321-1324.
[13]林小兰, 李传珍, 王晖, 等.基于贝叶斯理论的单音音乐节奏提取[J].中国传媒大学学报 (自然科学版), 2008, 15(4):44-51.
[14]石祥滨, 赵健谕, 刘芳, 等.一种分段式音乐情感识别方法[J].小型微型计算机系统, 2012, 33(8):1847-1850.
[15]Schuller B, Hage C, Schuller D, et al.‘Mister DJ, Cheer Me Up!′:Musical and textual features for automatic mood classification[J].Journal of New Music Research, 2010, 39(1):13-34.
[16]彭琼.上海音乐情感的计算机分析与自动识别技术研究[D].上海:上海交通大学, 2007.
