摘要
针对目前噪声环境下电子耳蜗汉语声调和语句识别率急剧下降问题,提出了一种基于时间精细结构过零刺激方案(FSZC)。利用语音信号时间精细结构的过零点时刻产生高速脉冲刺激序列,从而不仅将时间精细结构编码到电子耳蜗语音算法中,同时采用过零点时刻非均匀的采样脉冲刺激序列,确保了语音的空间编码和时间编码。声学模拟实验表明,相比于连续交替采样策略(CIS)和过零点时刻刺激策略(ZCT),FSZC方案在汉语声调和语句识别率上都有较大程度的提高,而且在噪声环境下对于声调和语句识别具有相对较强的抗噪性。
The performance of Mandarin tone and sentence recognition in cochlear implant falls sharply in noise environment. In order to solve the problem, this paper proposes a coding strategy based on the Zero-Crossing time of the temporal Fine Structure(FSZC)of the speech. The high-speed pulse stimulation sequences are generated at the zero-crossing time of the fine structure, thus the coding strategy not only includes the fine structure but also uses the non-uniform pulses which ensures the spatial and time encoding of the speech. Acoustic simulation results show that compared with Continuous Interleaved Sampling(CIS)strategy and the Zero-Crossing Time(ZCT)stimulating strategy, the Mandarin tone and sentence recognition rate have improved to a greater extent with the FSZC strategy, and the FSZC strategy has a stronger anti-noise performance.
出处
《计算机工程与应用》
CSCD
2014年第19期232-236,共5页
Computer Engineering and Applications
基金
国家自然科学基金(No.61271359)
苏州大学捷美生物医学工程仪器联合重点实验室资助项目
关键词
时间精细结构
过零点
汉语声调
电子耳蜗
temporal fine structure
zero-crossing
Mandarin tone
Cochlear Implant(CI)