Unterdrückung hörbarer Störgeräusche in Echtzeitsystemen

Diploma thesis (pdf 2,77MB)

To avoid distortion and processing delay in a hands free communication system, this work uses an auditory Gammatone filter bank instead of a fourier transform (FFT). The filter bank can be fully implemented in the time domain and thus doesn't need block processing. Additional singal buffering can be omitted completely, which helps in reducing processing delay. The major advantage of auditory signal analysis is that it contains pitch perception and simultaneous masking of the human ear inherently. Temporal pre and post masking models can be achieved with simple additional effort. As a consequence, the Gammatone filter bank based noise reduction and noise estimation only reconstructs the masking threshold of the undisturbed signal, so its effect is reduced to the suppression of audible noise. In this way the influence on the signal is kept as low as possible and musical noise is further reduced.