作　　者： ; ;

机构地区： 茂名学院计算机与电子信息学院

出　　处： 《信息技术》 2009年第8期68-70,73,共4页

摘　　要： 系统通过提取线性预测倒谱系数作为特征参数进行说话人识别研究。在具有VLIW(甚长指令集)体系结构的DSP芯片上,对提取的特征参数在进行并行遗传算法寻优,以获取说话人的最优参数进行说话人识别。实验结果表明,采用硬件并行遗传算法比简单遗传算法对特征参数进行优化识别,识别时间大幅减少,识别率有较大提高,为说话人识别的应用提供较强的参考价值。 The system researches speaker recognition by extracting linear prediction cepstnnn coefficient as characteristic parameters. To the extracted characteristic parameters, the system uses the parallel genetic algorithm optimization to obtain optimal parameters of speaker for speaker recognition on the VLIW architecture DSP chip. The experimental results show that using the hardware with parallel genetic algorithm to optimize the parameters of the characteristics for recognition, recognition time is reduced significantly and recognition rate is improved obviously as compared with the simple genetic algorithm, it can provide a strong reference value for speaker recognition.

关 键 词： 说话人识别 并行遗传算法 线性预测倒谱系数

领　　域： [自动化与计算机技术] [自动化与计算机技术]