基于汉语端点检测的电子耳蜗语音增强系统  

Speech enhancement system based on Chinese endpoint detection for cochlear implant

作　　者：卫铭斐[1] 杨放[1] 江莉[1] 许捷[1] WEI Ming-fei;YANG Fang;JIANG Li;XU Jie(School of Information and Control,Xi′an University of Architecture and Technology,Xi’an 710055,China)

机构地区：[1]西安建筑科技大学信息与控制工程学院,陕西西安710055

出　　处：《电子设计工程》2019年第4期109-112,117,共5页Electronic Design Engineering

基　　金：陕西省教育厅专项科研计划项目(15JK1432)

摘　　要：电子耳蜗语音处理一般是针对无声调的西方语设计的,随着电子耳蜗汉语使用者人数的快速增长,提高电子耳蜗汉语识别能力是面临的重要问题。将汉语端点检测和高级混合编码(ACE)策略相结合,利用基于元音的端点检测获取的静音段、特殊噪音段和音节端点等信息优化ACE策略,控制通道选择,调整刺激速率,可以增强汉语语音识别能力,更好地抑制含噪语音的噪声成分。通过主客观评价结果表明,基于汉语端点检测改进的ACE策略构成的电子耳蜗系统,相比传统CIS策略和ACE策略的电子耳蜗系统,平均识别率分别平均提高13%和11%,有效提高了电子耳蜗合成语音的清晰度和辨识度,增强了汉语语音识别效果。The speech processing of cochlear implant is generally designed for Western language with no tone. With the rapid growth of the number of Chinese users in cochlear implant,it is an important problem to improve the Chinese recognition ability of cochlear. By combining Chinese endpoint detection and advanced combination encoding(ACE)strategy,the ACE strategy can be optimized based on static segment, special noise segment and syllable endpoint information obtained from vowel endpoint detection. Those information control the channel selection and adjust the stimulus rate,which can enhance the power of Chinese speech recognition and better suppress the noise components of the noisy speech. Through subjective and objective evaluation,the results show that the cochlea implant system of the improved ACE strategy based on the Chinese endpoint detection,compared with the traditional CIS strategy and ACE strategy,the average recognition rate increased by 13% and 11% respectively,which effectively improved the clarity and recognition of synthesized speech in the cochlea implant and enhanced the Chinese speech recognition effect.

关 键 词：电子耳蜗 端点检测 语音增强 ACE策略 汉语端点 语音识别 

分 类 号：TN912[电子电信—通信与信息系统]