基于神经网络的5种音频分类

5 types Audio Classification Based on Nueron Networks

本文将音频信号的MFCC参数作为特征向量,并使用前馈型神经网络对语音、音乐、语音+音乐、环境音响、静音5类音频进行分类,取得了平均92%的正确率。

In this paper,MFCC is abstracted as feature vectors of the audio signals,and the Nueron Networks is chosen to classify five audio documents：speech, music, speech＋music, background and silence. The experimental results show that Nueron Networks is excellent for classification of audio documents,and the average classification accuracy is up to 92%.