Algorithmen in Akustik und Computermusik 2 - Seminar

Automatisierte Lautsprecherentzerrung mit Kautz-Filtern

A Kautz-filter cascade can be used to implement a customized frequency resolution for equalization of loudspeaker responses. This is particularly interesting for automatic eqalization as no (third-)octave filter bank has to be designed and manually adjusted. According to a paper from Paatero and Karjalainen, it is sufficient to design the frequency resolution, and use the filter structure with the impulse responses to solve normal equations. This seminar paper shows the mathematical formulation and practical results using a minimum-phase conversion.

Single-Channel Suppression of Background Noise in Speech Signals

This seminar paper shows the basics of background-noise suppression using the theory of spectral subtraction (Virag) and using either DFT or an auditory gammatone-filterbank (Slaney). The results from the uniform block-DFT based approach is further compared to the results obtained by the time-domain auditory filterbank approach. Moreover, the paper shows how to estimate background noise and eliminate musical-noise in post-processing.

Hyperinterpolation zur Abhöre des Direktschalls von Musikinstrumenten an frei wählbarer Abstrahlungsrichtung

Musical instruments sound differently in different directions of radiating sound. The presented seminar paper demonstrates an algorithm to listen to the recordings of a 64 channel surrounding spherical microphone array that records 64 directions of radiating sound simultaneously. In the Pure Data implementation of the presented algorithm using hyperinterpolation, a variable virtual microphone position that is not restricted to the 64 recording positions can be freely chosen by the user. See also