期刊论文

Speech/Music Discrimination with High Accuracy Based on Deep Belief Network

中文

吉首大学学报(自然科学版)

1007-2985

2017

38

1

36-39

10.3969/j.cnki.jdxb.2017.01.008

(61471164,61605045):国家自然科学基金资助项目 (2015JZ30):湖南省益阳市科技计划项目

本校为第一机构

运用深度置信网络进行语音/音乐识别,根据语音段中高频能量低于音乐段中高频能量的特点,对音频能量谱的高频部分进行学习,从而判别语音段或音乐段,MATLAB仿真实验结果验证了该方法的有效性和准确度.

Application of deep belief network in speech/music discrimination is studied. According to the characteristics that medium-high frequency energy of speech is lower than that of music, the high fre- quency of energy spectrum is examined to discriminate speech segment or music segment. MATLAB simulation results show the ef...