A time-frequency auditory model is presented. The model uses the wavel
et packet analysis as the preprocessor. The auditory filters are model
ed by the rounded exponential filters and the excitation is smoothed b
y a window function. By comparing time-frequency excitation patterns i
t is shown that the change in the time-frequency excitation pattern in
troduced when a test tone at masked threshold is added to the masker i
s approximately equal to 7 dB for all types of maskers. The classic de
tection ratio therefore overrates the detection efficiency of the audi
tory system.