Impulse response estimation and analysis in a shoebox room using MLS

Sammanfattning

This work investigates simultaneous, multi-source impulse response estimation in a simulated shoebox room using cross-correlation with Maximum Length Sequence (MLS) signals, deconvolution with swept-sine excitation, and least-squares (LS) inversion with MLS signals. The influence of MLS order and microphone position on estimation accuracy is also studied, as well as the possibility of simultaneous multi-source estimation in the presence of one, two and three simultaneous sources. All methods are evaluated for the single-source scenario. For multi-source estimation, cross-correlation and LS inversion are evaluated for two simultaneous sources, while LS inversion is further evaluated for three-source scenarios. The performance is assessed using the relative normalized mean-square error (RNMSE). Overall, the results show that LS-based estimation provides the highest accuracy, while MLS-based cross-correlation performs well for single-source scenarios. The estimation accuracy is influenced by factors such as MLS order and microphone position. For multi-source estimation, two simultaneous sources can be separated with good accuracy using LS inversion. However, the performance degrades significantly as the number of sources increases, and in the three-source case the estimation becomes unstable, with errors increasing by several orders of magnitude. These findings highlight the importance of excitation signal design and show that multi-source impulse response estimation becomes increasingly challenging as the number of simultaneous sources increases.