The speech stream simultaneously carries information about talker identity and linguistic content, and the same acoustic property (e.g., voice-onset-time, or VOT) may be used for both purposes. Separable neural networks for processing talker identity and phone... ...
