基于Gamma tone滤波和PCNN的说话人识别

为了更好地提取说话人的特征，对语音进行滤波的预处理。Gamma tone滤波器可以很好模拟基底膜滤波器的形状，因此本文采用作为标准耳蜗听觉滤波器的Gamma tone滤波器对语谱图进行滤波，再通过脉冲耦合神经网络来提取说话人特征进行分析。仿真结果表明，由该方法得到的说话人的熵序列与自身熵序列的欧氏距离较未经过Gamma tone滤波的小，提高了识别率。

Speaker recognition based on Gamma tone filter and PCNN

In order even better to extract feature parameters of speakers, speech should be filtered as a pre-processing process. Gamma tone filter can well imitate the human basilar membrane's shape,so this paper adopts Gamma tone filter which is a standard cochlea sense of hearing filter to filter spectrogram, then extracting speaker's characteristics through pulse coupled neural network(PCNN). Simulation results show that Euclidean distance of the speaker's entropy sequences through Gamma tone filter is shorter than that not through Gamma tone filter, which improves the recognition rate.