Article ID: 2025EAP1175
Room impulse responses (RIRs) between a loudspeaker and multiple positions characterize acoustic propagation in a room. RIRs are essential for spatial acoustic applications such as sound field control and visualization. However, measuring RIRs over a wide region requires dense spatial sampling, which demands large-scale microphone arrays or long measurement times. Consequently, this limitation has motivated research on spatial interpolation and extrapolation methods for RIRs based on physical sound field models. The sparse equivalent source method (SESM) enables sound field estimation from limited measurements by modeling the field as a sparse superposition of equivalent sources. However, in SESM, the estimation accuracy deteriorates when the weights of many equivalent sources are estimated at once, because the sensing matrix becomes coherent. To overcome this limitation, this study proposes an estimation method for early RIRs that uses prior information on the arrival times of the reflections to decompose the early RIR signals in the time domain, thereby dividing the single large inverse problem into smaller subproblems, each using only the equivalent sources grouped around the corresponding image sources. This decomposition reduces both the size and the coherence of the sensing matrix, making the optimization more stable and improving the estimation accuracy of equivalent source weights. Simulation and measurement experiments were conducted to evaluate the proposed method. The simulations demonstrated that the SESM with grouped image sources improved estimation accuracy compared with conventional SESM, particularly for higher reflection orders, and expanded the region with high estimation accuracy. In addition, experimental results in an anechoic chamber with reflecting panels confirmed that the proposed method outperformed the conventional SESM in terms of estimation accuracy.