【作者】 冯朝斌； 吕成国； 赵洪刚；

【Author】 FENG Chao-bin,LV Cheng-guo,ZHAO Hong-gang (School of Computer Science and Technology,Heilongjiang University,Harbin 150080,China)

【机构】 黑龙江大学计算机科学与技术学院；

【摘要】 在说话人识别系统中,系统的响应速度是影响系统性能的一个重要的因素。为了提高说话人识别系统的响应速度,将说话人模型间距离测算的方法引入到系统的识别过程之中。在距离测算方法的基础上,提出了一种方法在说话人系统识别之前进行预处理,通过实验证明这种方法可以在不明显降低识别率的情况下有效地提高系统的执行效率。另外,将拒识策略加入到系统之中,它可以将可信度不高的结果过滤掉,从而提高系统的识别率。实验结果显示,这两种方法的结合可以有效的提高系统的性能。更多还原

【Abstract】 The system response speed is an important factor of the speaker recognition system performance.In order to improve the speaker recognition system response speed,the distance between the speaker model estimation methods was introduced.From the basis of calculation method,a method that pre-treatment before recognition was introduced. Experiments proved that this method can effectively improve system performance.In addition,the rejection strategy in this article will be added to the system,and it credibility can filter out the results whose credible was very low,thereby enhancing the system recognition rate.Experimental results showed that the combination of these two methods can effectively improve the performance of the system.更多还原