全局谱参数下的耳语说话人状态因子分析

发布时间：2018-05-17 02:16

本文选题：耳语音 + 说话人识别　；参考：《声学学报》2014年02期

【摘要】：提出了全局谱参数下的耳语说话人状态因子分析方法。首先,根据耳语听辨实验结果,提出导入唤醒度-愉悦度因子对说话人状态进行三级度量;其次,提取耳语音正弦模型、人耳听觉模型下的谱参数,结合其他短时频谱参量,进行轨迹跟踪并计算各参数的全局统计变量,作为特征参数来实现耳语说话人状态的分类。实验结果显示,正弦模型及人耳听觉模型的全局谱参数可将耳语说话人状态因子分类系统的准确率提高至90%。该分类方法及状态因子描述方案提供了耳语音说话人状态分析的有效途径。
[Abstract]:In this paper, a method of whispering speaker state factor analysis based on global spectral parameters is proposed. First of all, according to the experimental results of hearing in whispers, we propose a three-level measurement of the speaker's state by introducing the arousal and pleasure factors. Secondly, we extract the spectral parameters of the sinusoidal model and the auditory model, and combine with other short-term spectral parameters. Track and calculate the global statistical variables of each parameter, as characteristic parameters to achieve the classification of whispers' states. The experimental results show that the global spectral parameters of sinusoidal model and auditory model can improve the accuracy of the classification system of whispering speaker state factors to 90. The classification method and the description scheme of the state factor provide an effective way to analyze the state of the speaker.
【作者单位】：苏州大学电子信息学院;苏州大学物理科学与技术学院;
【基金】：国家自然科学基金(61071215,61271359,61372146) 江苏省普通高校研究生科研创新计划项目(05KJB510113)资助
【分类号】：TN912.3

【参考文献】