目前在说话人识别中常用的特征是线性预测编码(LPC)参数和美尔倒谱系数(M FCC)等。在说话人识别系统中我们首先使用美尔倒谱系数作为参数,然后将美尔倒谱系数和L em pe l-Z iv复杂性相结合。实验结果显示,在50个人的文本相关说话人识别实验中,识别率从42%提高到80%;在50个人的文本无关说话人识别中,识别率从60%提高到72%。提示L em pe l-Z iv复杂性作为一种新的特征参数可以应用于说话人识别。 At present, the commonly used features in speaker recognition are Linear Predictive Coding (LPC) parameters and Mercard-Cepstral Coefficients (M FCC). In the speaker recognition system, we first use the cepstral coefficients as parameters, and then combine the celestial cepstral coefficients with the Limpel-Ziv complexity. The experimental results show that the recognition rate increases from 42% to 80% in 50 people’s text-dependent speaker recognition experiments, and from 60% to 72% in 50 people’s text-independent speaker recognition. It is suggested that the complexity of L em pe l -Z iv can be applied to speaker recognition as a new feature parameter.