A methodology is presented for simulating with an auditory model some psychoacoustic properties of a human listener. Masking thresholds were measured for a listener considered an expert in audio-quality assessment. Then, several parameters of the auditory model were adapted using a genetic algorithm so that the model's detection thresholds matched the listener's thresholds. The resulting model showed improved performance in predicting mean audio-quality ratings obtained from tests with human listeners.
This paper costs $33 for non-members and is free for AES members and E-Library subscribers.