Signal adaptive auto tuning algorithm for microphone-loudspeaker systems using empirical mode decomposition

Speaker tuning is the process of adjusting the frequency response of an audio system to ensure clear, balanced sound reproduction. Improper tuning can lead to unintended tonal distortion or uncontrolled acoustic loops. Traditionally, tuning requires skilled engineers to manually configure Parametric Equalizers (PEQs) for each loudspeaker–microphone pair, which is time-consuming and difficult to scale. Existing automated approaches, such as inverse filtering or peak detection, often suffer from noise sensitivity, spatial inconsistency, or overcompensation artifacts. We propose a novel, signal-adaptiveauto-tuning algorithm that eliminates the need for manual parameter adjustment. Unlike conventional high-pass filters, which require manually set cutoff frequencies and Q-factors, our method uses Empirical Mode Decomposition (EMD) to adaptively isolate high-frequency, feedback-prone components of the system response. The resulting response is equalized using a Minimum Phase Filter (MPF) implemented as a Finite Impulse Response Equalization (FIR-EQ) filter. Objective evaluation across multiple loudspeaker–microphone configurations and reverberant conditions shows that our method reduces global spectral kurtosis by 0.34 and moving-octave-band kurtosis by 0.35, while achieving the greatest improvement in spectral flatness among all compared methods. Our method preserves tonal naturalness and operates in less than 200 ms, making it suitable for automatic speaker tuning in diverse acoustic environments.