Robust Listening Room Compensation by Optimizing Multiple p-Norm Based Criteria

The purpose of room impulse response reshaping is to reduce reverberation and thus to improve the perceived quality of the received signal by prefiltering the source signal before it is played with a loudspeaker. The optimization of an infinityand/or p-norm based objective function in the time domain has shown to be quite effective compared to least-squares methods. Multi-position approaches have been developed in order to increase the robustness against small movements of the listener. It has been shown recently, that it is necessary to also consider the frequency-domain representation of the reshaped impulse response, in order to avoid spectral distortions of the overall system. In this contribution we propose to jointly optimize a p-norm based criterion in the time and frequency domains for multiple positions in order to achieve a good reshaping while not affecting the perceived quality due to spectral distortions in a limited listening area. Multi-Point Reshaping by p-Norm Optimization By designing reshaping filters, one aims at improving the acoustic properties of a channel. Early approaches tried to invert the acoustic channel. Recent methods exploit psychoacoustic properties of the human auditory system to lighten the pressure on the equalized overall system: echoes may occur if they are below the average temporal masking limit. Furthermore, the optimization of a pnorm based optimality criterion has proven to result in a superior overall reshaping compared to least-squares approaches. One problem is the lack of spatial robustness as RIRs are very sensible, even to small movements. If the loudspeaker or the listener move slightly, the perceived quality is degraded. An efficient implementation of a multipoint equalization scheme utilizing multiple loudspeakers is given in [2]: according to the spatial sampling theorem of RIRs, it is sufficient to design the reshaping filters for a finite number of sampled RIRs in a small volume to perform reshaping inside the whole area [2]. Considering a setup consisting of N loudspeakers and R RIRs sampled in a finite area, one aims to design one prefilter for each loudspeakers to jointly reshape all of the measured RIRs in the listening area. Let c (r) (n) denote the r-th sampled RIR from loudspeaker to the listening area, whereas the prefilter for the -th loudspeaker is denoted as h (n). The global impulse response (GIR) at the r-th sampling point is given by