Abstract: A fast classification method of narrow-band audio signals is proposed based on the linear discriminate analysis and Gaussian mixture model.The speech，music and noise signals can be effectively discriminated with this method in the environment of white noise，street noise and car interior noise respectively.The experimental results show that the proposed method can make the classification accuracy close to 95% while classification time is less than one second.

Key words: fast audio classification, feature extraction, Linear Discriminant Analysis（LDA）, Gaussian Mixture Model（GMM）

摘要： 提出了一种基于线性判别分析和高斯混合模型的窄带音频快速分类方法，该方法在白噪声、街道噪声和车内噪声环境下都能有效区分语音、音乐和噪声。实验结果表明，该方法在保证分类时间不大于1 s的情况下，分类准确率能达到95%以上。

关键词: 音频快速分类, 特征提取, 线性判别分析, 高斯混合模型