【作者】 沈燕；

【导师】 顾济华； 陶智；

【作者基本信息】 苏州大学 ， 检测技术与自动化装置， 2012， 硕士

【摘要】 电子耳蜗是一种利用功能性电刺激直接兴奋听神经来恢复耳聋患者的听觉感知的装置，但是目前针对汉语音声调的语音编码策略还存在不足，如声调感知发音效果较差，对旋律和音乐感知较弱。针对电子耳蜗植入者对声调感知的不足，本文研究了时域信息与频域信息的获取率对声调感知的决定性作用，并提出了基于听觉感知的小波包分解策略，采用多分辨率分析思想，非均匀划分时频空间，获取有效的时频信息，从而提高患者音调和音质的感知能力。针对耳蜗植入电极数目的限制，在保证完整获取时频域信息和避免多电极通道相互间干扰的基础上，采用通道能量最大选取准则，剔除语音信息含量较少的频率通道，即保证了有效时频信息的传递，又避免了电极通道间的相互干扰。针对基于高速脉冲调制时产生的带内噪声等对汉语声调的掩蔽干扰，提出了调制深度选择方案，极大的削弱了带内噪声的干扰，进而提高了汉语声调信息的获取率。研究结果表明，基于听觉感知的小波包分解与调制深度选择相结合的策略，能有效地抑制带内噪声，并获取更丰富的时频信息，增强患者声调感知能力。更多还原

【Abstract】 Cochlear implant (CI) is a medical device that can restore the residual auditoryperception for the profoundly or totally deafened person via functional electricalstimulation. At present, however，the speech coding strategies for Chinese tone still havesome shortcomings, such as poor perception of tone, melodies and music.According to the patients’ lack of tone perception, based on the decisive role ofinformation retrieval to tone perception in both time domain and frequency domain, wepropose the wavelet packet decomposition strategy using multi-resolution analysis andnon-uniformity dividing time-frequency space technology. To effectively accesstime-frequency information in order to improve the patients’ tone pitch perception andspeech quality.On the account of the limited number of cochlear implant electrode, basing onensuring the complete information obtained from the time-frequency domain andpreventing the multi-electrode channel mutual interference, we apply the channelmaximum energy criterion for selection of the effective channels so as to remove thechannels containing the less information of speech, which is to ensure the transmission ofeffective information and to avoid interference between the electrode channels.On account of the noise masking interference of Chinese tones causing by thehigh-speed pulse modulation, we propose the modulation depth options which greatlyweak the interference of in-band noise and thus improve the access to Chinese tones.The results show that the combination of wavelet packet decomposition basing onauditory perception and depth of modulation can effectively suppress the in-band noiseand enhance the tone perception.更多还原