Nagarajan Srikantan S, Cheung Steven W, Bedenbaugh Purvis, Beitel Ralph E, Schreiner Christoph E, Merzenich Michael M
Department of Bioengineering, University of Utah, Salt Lake City, Utah 84112-9458, USA.
J Neurophysiol. 2002 Apr;87(4):1723-37. doi: 10.1152/jn.00632.2001.
Cortical sensitivity in representations of behaviorally relevant complex input signals was examined in recordings from primary auditory cortical neurons (AI) in adult, barbiturate-anesthetized common marmoset monkeys (Callithrix jacchus). We studied the robustness of distributed responses to natural and degraded forms of twitter calls, social contact vocalizations comprising several quasi-periodic phrases of frequency and AM. We recorded neuronal responses to a monkey's own twitter call (MOC), degraded forms of their twitter call, and sinusoidal amplitude modulated (SAM) tones with modulation rates similar to those of twitter calls. In spectral envelope degradation, calls with narrowband channels of varying bandwidths had the same temporal envelope as a natural call. However, the carrier phase was randomized within each narrowband channel. In temporal envelope degradation, the temporal envelope within narrowband channels was filtered while the carrier frequencies and phases remained unchanged. In a third form of degradation, noise was added to the natural calls. Spatiotemporal discharge patterns in AI both within and across frequency bands encoded spectrotemporal acoustic features in the call although the encoded response is an abstract version of the call. The average temporal response pattern in AI, however, was significantly correlated with the average temporal envelope for each phrase of a call. Response entrainment to MOC was significantly correlated with entrainment to SAM stimuli at comparable modulation frequencies. Sensitivity of the response patterns to MOC was substantially greater for temporal envelope than for spectral envelope degradations. The distributed responses in AI were robust to additive continuous noise at signal-to-noise ratios > or =10 dB. Neurophysiological data reflecting response sensitivity in AI to these forms of degradation closely parallel human psychophysical results on the intelligibility of degraded speech in quiet and noisy conditions.
在成年巴比妥麻醉的普通狨猴（Callithrix jacchus）初级听觉皮层神经元（AI）的记录中，研究了行为相关复杂输入信号表征中的皮层敏感性。我们研究了对自然和退化形式的啁啾叫声（由几个频率和调幅的准周期短语组成的社交接触发声）的分布式反应的稳健性。我们记录了神经元对猴子自身啁啾叫声（MOC）、其啁啾叫声的退化形式以及调制率与啁啾叫声相似的正弦调幅（SAM）音调的反应。在频谱包络退化中，具有不同带宽窄带通道且与自然叫声具有相同时间包络的叫声，其载波相位在每个窄带通道内是随机的。在时间包络退化中，窄带通道内的时间包络被滤波，而载波频率和相位保持不变。在第三种退化形式中，向自然叫声中添加噪声。尽管编码反应是叫声的抽象版本，但AI内和跨频段的时空放电模式对叫声中的频谱时间声学特征进行了编码。然而，AI中的平均时间反应模式与叫声每个短语的平均时间包络显著相关。在可比调制频率下，对MOC的反应夹带与对SAM刺激的夹带显著相关。反应模式对MOC的敏感性在时间包络方面比对频谱包络退化要大得多。当信噪比≥1０dB时，AI中的分布式反应对加性连续噪声具有稳健性。反映AI对这些退化形式反应敏感性的神经生理学数据与人类在安静和嘈杂条件下对退化语音可懂度的心理物理学结果密切平行。
