噪声环境下基于能量检测的生态声音识别

摘要

无效声音段的存在导致噪声环境下声音识别方法的识别率迅速降低.为此,提出一种基于能量检测的抗噪声音识别方法.用能量检测方法检测背景噪声环境下的有用声音信号,对检测到的信号提取Mel频率倒谱系数特征,使用支持向量机对提取的特征向量训练分类模型,对含有噪声的生态环境声音进行识别.实验结果表明,该方法具有较好的抗噪能力,其在信噪比40 dB以下的识别率比添加能量检测前提高约25％.%The presence of invalid sound segments leads to quickly reduce the recognition performance of the sound recognition method in the noise condition. This paper presents an anti-noise sound recognition method based on the energy detection. Energy detection method is used to detect the existence of the sound signal under the noise conditions. In addition, it extracts Mel-frequency Cepstral Coefficient(MFCC) features from the frames which are detected the presence of the sound signal by the energy detection. A Support Vector Machine(SVM) classifier uses the extracted feature vectors to train the classification model and recognizes the ecological environmental sound. Experimental results show that this method has good anti-noise ability, and the recognition rates are more than 25% higher than the system before adding the energy detection when the Signal to Noise Ratio(SNR) is less than 40 dB.