Dillier N, Bögli H, Lai W K
ENT Department, University Hospital Zürich, Switzerland.
Ann Otol Rhinol Laryngol Suppl. 1995 Sep;166:354-6.
The Speak speech-coding strategy for the Nucleus Minisystem-22 cochlear implant continuously analyzes the speech signal using 20 digitally programmable band-pass filters and presents up to 10 spectral maxima to the implanted electrodes. To analyze the performance of this system for a variety of speech sounds in quiet and noise, the stimulation patterns of the implanted electrode array were reconstructed from the transmitted radio frequency signals by software as electrodograms and compared to electrodograms generated by other speech-coding strategies, as well as to the spectrograms of the input signals. The performance with the Speak strategy relative to that with the Multipeak (Mpeak) speech-processing strategy was also evaluated in a field trial study with 20 native German-speaking cochlear implant users from four European implant centers, involving a variety of auditory perceptual tasks in an ABAB paradigm over a 12-week period. Vowel, consonant, and monosyllable word tests, as well as sentence tests in quiet and noise, were conducted. Significant differences in group mean scores for most speech recognition subtests were obtained for the Speak versus the Mpeak strategy, with the largest overall improvements observed for the sentence tests in noisy conditions.
核子迷你系统22型人工耳蜗的Speak语音编码策略使用20个数字可编程带通滤波器持续分析语音信号，并将多达10个频谱最大值发送到植入电极。为了分析该系统在安静和噪声环境中对各种语音声音的性能，通过软件将植入电极阵列的刺激模式从传输的射频信号重建为电描记图，并与其他语音编码策略生成的电描记图以及输入信号的频谱图进行比较。在一项针对来自四个欧洲植入中心的20名以德语为母语的人工耳蜗使用者的现场试验研究中，也评估了Speak策略相对于多峰（Mpeak）语音处理策略的性能，该研究在12周内采用ABAB范式涉及各种听觉感知任务。进行了元音、辅音和单音节词测试，以及安静和噪声环境下的句子测试。对于Speak策略与Mpeak策略，大多数语音识别子测试的组平均分数存在显著差异，在噪声环境下的句子测试中观察到总体最大的改进。
