首页> 外国专利> Speaker judgment device, speaker determination information generation method and program

Speaker judgment device, speaker determination information generation method and program

机译:说话者判断装置,说话者确定信息生成方法和程序

摘要

To provide a speaker determination device which is improved in speaker determination precision.SOLUTION: A speaker determination device includes: a similarity calculation part which calculates similarity between a speaker feature quantity of each divided speech section obtained by dividing a speech section of a speech signal according to a predetermined time length and a speaker feature quantity previously generated for each person in charge of window work; a speaker primary determination part which generates primary determination information representing a speaker ID of each divided speech section from the similarity; a speaker secondary determination part which generates secondary determination information by using a speaker ID of a nearby speaker as secondary determination information of an arbitrary divided speech section when similarity between a speaker feature quantity of the nearby speaker as the most matching speaker in a predetermined number of divided speech sections before or after the arbitrary divided speech section and a speaker feature quantity of the arbitrary divided speed section meets predetermined conditions; and a speaker clustering part which generates a speaker ID of a customer by clustering speaker feature quantities of divided speech sections corresponding to the secondary determination information representing the customer, i.e., a set of customer speaker feature quantities so as to generate tertiary determination information.SELECTED DRAWING: Figure 1
机译:为了提供一种提高说话者确定精度的说话者确定装置。解决方案:说话者确定装置包括:相似度计算部分,该相似度计算部分计算通过根据语音信号的语音部分进行划分而获得的每个划分的语音部分的说话者特征量之间的相似度。预先确定的时间长度和先前为每个负责窗口工作的人生成的说话者特征量;扬声器主要确定部分,其根据相似度生成表示每个划分的语音区间的讲话者ID的主要确定信息;扬声器次要确定部分,当预定数量的附近作为最匹配扬声器的讲话者的讲话者特征量之间的相似性时,通过使用附近讲话者的讲话者ID作为任意划分的语音区间的次要确定信息来生成次要确定信息。任意分割语音区间的前后的分割语音区间,以及任意分割速度区间的说话者特征量满足规定条件。扬声器聚类部分,其通过聚类对应于代表顾客的次级确定信息的划分语音区间的扬声器特征量,即一组顾客扬声器特征量,来产生顾客的扬声器ID,以产生第三判定信息。图:图1

著录项

相似文献

  • 专利
  • 外文文献
  • 中文文献
获取专利

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号