| 摘要 |
A method for detecting, in an audio signal including a stream of audio samples (y(n), (y′(n)), ticks in a noisy environment, comprises the steps of applying the signal to a coarse tick detection processor arranged to decide whether it is likely that the signal includes a tick and, only then, enabling a fine tick detection processor to decide, by more thorough processing the audio signal, whether the audio signal represents an audio tick. The coarse processing step preferably includes buffering audio samples (y(n), (y′(n)), determining the maximum and minimum values of each sample, forming their local range, delaying the local range with one or more samples and comparing the difference between the actual local range and the delayed local range with a threshold value (Tc), and, if the difference exceeds the threshold value, outputting a trigger (nc) for enabling the fine detection processor The fine processing step preferably includes buffering audio samples (y(n), (y′(n)), computing, for each buffer content (Bp(n,m)), a Fast Fourier Transformed buffer content (Ec(ω,m)), determining the difference (Fc(ω,m)) between the resulting frequency components, comparing the difference (Fc(ω,m)) to the content of a previously trained fingerprint set (F*(ω,m)) and calculating a correlation coefficient (p(Fc,F*)) between them, comparing the correlation coefficient to a threshold value (Tp) and, if the correlation coefficient exceeds the threshold value, outputting a signal indication the presence of a tick in the audio signal.
|
