基于DTW的音乐信号特征识别系统研究
首发时间:2021-08-05
摘要:以音乐信号的特征提取与识别为研究背景,提出基于DTW的音乐信号特征识别系统。该系统采用信号预加重和信号分帧的方法,并结合汉明窗函数对信号进行加权处理。通过检测短期能量和检测短期过零率实现双门限端点检测。将采用傅里叶变换求得的功率谱通过滤波器组后取对数得到每帧的MFCC系数。采用DTW算法实现该音乐识别系统的识别与训练模式。仿真结果表明,采用本文的音乐识别系统可以有效的消除信号中的噪音,准确的检测出语音片段的端点,进行音乐的识别与重现。
关键词: 音乐识别系统 预处理 端点检测 MFCC特征参数 DTW?????
For information in English, please click here
Research on music Signal Feature recognition System based on DTW
Abstract: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.
Keywords: key word Music recognition system Pretreatment Endpoint detection MFCC characteristic parameters DTW
基金:
引用
No.****
动态公开评议
共计0人参与
勘误表
基于DTW的音乐信号特征识别系统研究
评论
全部评论0/1000