摘要: 现有说话人识别系统对环境噪声及说话人声音变迁等干扰的鲁棒性较差。为此,在改进和优化高斯混合-通用背景模型的基础上,根据家庭环境中的典型特征,设计并实现一种用于家用机器人的说话人识别系统。应用结果表明,该系统具有较好的识别性能和较高的鲁棒性,适用于声控门禁和语音签到等领域。
关键词:
说话人识别,
家用机器人,
梅尔频率倒谱系数,
高斯混合模型,
通用背景模型
Abstract: Based on a home robot platform, this paper implements a classical speaker recognition algorithm: Gaussian mixed-universal background model algorithm. It also introduces the speaker recognition theory in the robot system. In order to improve the robustness of the system in real home environment, it make some improvements in framework and algorithms. The recognition system can be applied in access system or check-in system as well.
Key words:
speaker recognition,
home robot,
Mel Frequency Cepstrum Coefficient(MFCC),
Gaussian Mixed Model(GMM),
Universal Background Model(UBM)
中图分类号:
武宁, 肖星星, 冯瑞. 家用机器人的说话人识别系统[J]. 计算机工程, 2012, 38(2): 207-209.
WU Ning, XIAO Xing-Xing, FENG Rui. Speaker Recognition System of Home Robot[J]. Computer Engineering, 2012, 38(2): 207-209.