Xiu Zhai

The paper shows human speech recognition in natural noise relies on mid-level auditory statistics, with a model predicting perceptual variance better than cochlear models.

The paper shows that distinct neural response statistics in the auditory system encode sound texture features for recognition and discrimination, with fast spectrum-based discrimination and slow correlation-based recognition.

The paper shows that correlation statistics between frequency-organized auditory channels can be used to identify sound categories with high accuracy, mirroring human performance.