Geurts L, Wouters J
Laboratory for Experimental ORL, KULeuven, Belgium.
J Acoust Soc Am. 2001 Feb;109(2):713-26. doi: 10.1121/1.1340650.
In this study the perception of the fundamental frequency (F0) of periodic stimuli by cochlear implant users is investigated. A widely used speech processor is the Continuous Interleaved Sampling (CIS) processor, for which the fundamental frequency appears as temporal fluctuations in the envelopes at the output. Three experiments with four users of the LAURA (Registered trade mark of Philips Hearing Implants, now Cochlear Technology Centre Europe) cochlear implant were carried out to examine the influence of the modulation depth of these envelope fluctuations on pitch discrimination. In the first experiment, the subjects were asked to discriminate between two SAM (sinusoidally amplitude modulated) pulse trains on a single electrode channel differing in modulation frequency ( deltaf = 20%). As expected, the results showed a decrease in the performance for smaller modulation depths. Optimal performance was reached for modulation depths between 20% and 99%, depending on subject, electrode channel, and modulation frequency. In the second experiment, the smallest noticeable difference in F0 of synthetic vowels was measured for three algorithms that differed in the obtained modulation depth at the output: the default CIS strategy, the CIS strategy in which the F0 fluctuations in the envelope were removed (FLAT CIS), and a third CIS strategy, which was especially designed to control and increase the depth of these fluctuations (F0 CIS). In general, performance was poorest for the FLAT CIS strategy, where changes in F0 are only apparent as changes of the average amplitude in the channel outputs. This emphasizes the importance of temporal coding of F0 in the speech envelope for pitch perception. No significantly better results were obtained for the F0 CIS strategy compared to the default CIS strategy, although the latter results in envelope modulation depths at which sub-optimal scores were obtained in some cases of the first experiment. This indicates that less modulation is needed if all channels are stimulated with synchronous F0 fluctuations. This hypothesis is confirmed in a third experiment where subjects performed significantly better in a pitch discrimination task with SAM pulse trains, if three channels were stimulated concurrently, as opposed to only one.
在本研究中，对人工耳蜗使用者对周期性刺激的基频（F0）的感知进行了调查。一种广泛使用的语音处理器是连续交错采样（CIS）处理器，对于该处理器，基频在输出端表现为包络中的时间波动。对四名使用LAURA（飞利浦听力植入物的注册商标，现为欧洲耳蜗技术中心）人工耳蜗的用户进行了三项实验，以研究这些包络波动的调制深度对音高辨别力的影响。在第一个实验中，要求受试者在单个电极通道上区分两个在调制频率上不同（Δf = 20%）的正弦幅度调制（SAM）脉冲序列。正如预期的那样，结果表明调制深度较小时性能会下降。根据受试者、电极通道和调制频率的不同，调制深度在20%至99%之间时可达到最佳性能。在第二个实验中，针对三种在输出端获得的调制深度不同的算法，测量了合成元音F0的最小可察觉差异：默认的CIS策略、去除包络中F0波动的CIS策略（FLAT CIS）以及专门设计用于控制和增加这些波动深度的第三种CIS策略（F0 CIS）。一般来说，FLAT CIS策略的性能最差，在该策略中F0的变化仅表现为通道输出中平均幅度的变化。这强调了语音包络中F0的时间编码对音高感知的重要性。与默认的CIS策略相比，F0 CIS策略没有获得明显更好的结果，尽管后者在某些情况下会导致包络调制深度，在第一个实验中这些深度会得到次优分数。这表明如果所有通道都由同步的F0波动刺激，则需要的调制较少。这一假设在第三个实验中得到了证实，在该实验中，与仅刺激一个通道相比，如果同时刺激三个通道，受试者在SAM脉冲序列的音高辨别任务中的表现明显更好。
