Ambisonics decoder design based on sparse plane wave decomposition

Sound field reproduction by loudspeaker array is an important way to create acoustical virtual reality. High order ambisonics (HOA) is a highly applicable method to reproduce a desired sound field with a loudspeaker array. For a target sound field that is represented by a group of harmonic coefficients, the process of calculating loudspeaker weights for reproduction with these harmonic coefficients is called decoding. Mode-matching based decoding is usually applied in HOA reproduction and good performance is achieved in the loudspeaker array center (the ‘sweet spot’). However, the accuracy and sweet spot size are strongly affected by the recording order of the target sound field. To improve the reproduction performance, a new ambisonic decoding method is proposed based on the sparse plane wave decomposition in the case that the target sound field is steered by a small number of acoustic sources in this paper. The recorded target sound field is decomposed into a group of plane waves in the harmonic domain. Loudspeaker weights are calculated as a summation of the precalculated corresponding plane waves weights. Simulations are conducted in the free field and reverberation room to verify that the proposed method outperforms the mode-matching decoding. Experimental validation is conducted in a real listening room with a circular loudspeaker array. It is concluded that the reproduction error by the proposed method is superior to that by the sectorial mode matching for the investigated 2.5D reproduction case in the reverberant environment after 500 Hz and in terms of the MAC values, the proposed method ensures the spatial correspondence over frequency of the desired sound field and the reproduced sound field.