Today's automatic speech recognition has fundamental limits such as: isolated (single) speaker, speaker dependency, and limited vocabulary. Wave precedence, interaural delay, and masking can be combined in the time domain using simple logic, neural networks and fractal geometry to achieve Star Trek goals, i.e.: cocktail party processing, speaker independence, and phoneme recognition. Demonstrations and examples will illustrate these new synergistic principles.
This paper costs $33 for non-members and is free for AES members and E-Library subscribers.