Robust DOA Estimation Based on Adaptive Logarithmic Compression MUSIC

Impulsive acoustic interference can distort covariance estimates and destabilize direction-of-arrival (DOA) estimation with multiple signal classification (MUSIC). This study proposes adaptive logarithmic compression MUSIC (ALC-MUSIC) for acoustic arrays operating under time-varying burst interference. Its contribution is an adaptive compression framework that uses noise-subspace drift to detect statistical changes and tunes logarithmic compression parameters through a composite objective based on the Jarque–Bera statistic, skewness, and kurtosis. A dual-branch implementation combines projection approximation subspace tracking based on recursive least squares with exponentially weighted covariance estimation, supporting compression adaptation and MUSIC spectrum estimation. Experiments with a six-element MEMS microphone array tested a 1 kHz target under time-gated broadband Gaussian interference with burst widths of 50 and 100 ms. ALC-MUSIC achieved DOA root mean square errors of 3.1615° and 3.1478°, respectively, with 45.06% and 73.88% of estimates within ±2° of the target direction. The corresponding 90th-percentile absolute errors were 3.61° and 3.85°. These results demonstrate improved estimation stability and higher success rates than the tested MUSIC variants under the investigated burst-interference conditions, supporting adaptive amplitude compression for robust acoustic DOA estimation.