面向智能机器人的Teager语音情感交互系统设计与实现  

Design and implementation of speech emotion interaction system based on Teager for intelligent robot

作　　者：李翔[1] 李昕[1,2] 胡晨[1] 卢夏衍[1] 

机构地区：[1]上海大学机电工程与自动化学院,上海200072 [2]哈尔滨工业大学机器人技术与系统国家重点实验室,哈尔滨150001

出　　处：《仪器仪表学报》2013年第8期1826-1833,共8页Chinese Journal of Scientific Instrument

基　　金：机器人技术与系统国家重点实验室开放基金项目(SKLS-2009-MS-10)资助

摘　　要：自从Maragos首先提出了Teager能量算子(TEO)后,该算子得到了一系列应用。Teager能量算子的非线性特性,使之能在抑制背景噪声的同时进行信号特征的有效提取。它还保留了倒谱分析方法中的准稳态假设,因而更能有效地表征语音信号的复杂性。探索了基于Teager能量算子的非线性特征对于区分情感语音的有效性,并且进一步将其和普通的声学参数和音质参数相结合,以应用于语音情感的识别中。以HMM和GMM为分类器,选取7种情感设计了4组对比实验,达到最高85%的识别正确率。实验结果表明,将传统的语音情感特征和Teager能量算子的非线性特征相结合,能显著地提高语音情感识别系统的性能。在此基础上,提出了智能情感机器人进行情感交互的框架,设计实现了智能服务机器人的情感交互系统,经过实时情感交互实验验证,该智能服务机器人可以很好地与交互对象进行简单的情感交流。Since Maragos firstly put forward the Teager energy operator （TEO）, the operator has got a series of applications. The nonlinear characteristics of the Teager energy operator can effectively extract signal feature while inhibi- ting background noise at the same time. It also retains the quasi steady state assumption in cepstrum analysis method, and can more effectively represent the complexity of the speech signal. This paper discusses the validity of distinguis- hing emotional speech based on the nonlinear features of Teager energy operator, and combines the nonlinear features with traditional acoustic parameters and voice quality parameters, which are applied to speech emotion recognition. The HMM and GMM were used as the classifiers, seven kinds of emotion status were selected to design 4 groups of comparative experiments, and the highest recognition rate of 85% was achieved. The experimental results indicate that combining the nonlinear characteristic of the Teager energy operator with traditional emotional speech features can improve the performance of speech emotion recognition system remarkably. On this basis, this paper puts forward the framework of intelligent emotional robot making emotional interaction, designs and realizes the intelligent service robot emotional interaction system. Through real-time emotional interaction experiment verification, this intelligent service robot can make simple emotional communication with interactive objects.

关 键 词：TEAGER能量算子 非线性特征 智能机器人 语音情感交互系统 

分 类 号：TP391.42[自动化与计算机技术—计算机应用技术]