Effect of higher-order ambisonics on evaluating beamformer benefit in realistic acoustic environments

Multi-channel loudspeaker systems have been proposed to assess the real-life benefit of devices such as hearing aids, cochlear implants, or mobile phones. This paper investigates to what extent sound fields recreated by Higher-Order Ambisonics (HOA) can be used to evaluate the performance of spatially selective multi-microphone processing schemes (beamformers) inside complex acoustic environments. Two example schemes are considered: an adaptive directional microphone (ADM) and a contralateral suppression bilateral beamformer (BBF), both implemented in the context of a hearing aid device. The acoustic scenarios consist of a single speech target (0°) competing against three speech jammers (±90° and 180°) set either in an anechoic or in a reverberant simulated classroom (T₃₀ = 0.6s). The HOA effect on the directional algorithm performance is quantified through: (a) the adaptive, frequency-dependent, algorithm gains, (b) the SNR improvement calculated in one-third octave bands, and (c) the processed target frequency response. The HOA reconstruction errors influence the beamformers in mainly two ways; first, by altering the spatial characteristics of the sound field, which in turn modifies the adaptation of the algorithms, and second, by affecting the spectral content of the sources. The results suggest that although HOA (here 7^th order) does not degrade the broadband, long-term, intelligibility-weighted SNR improvement of the two beamformers, it imposes a low-pass effect on the processed target. This renders the HOA coding problematic above the system's cut-off frequency.