Sampling Frequency Mismatch Estimation by Auxiliary-Function-Based Iterative Maximization of Double-Cross-Correlation

Abstract

In this paper, we propose a new variant of sampling frequency mismatch (SFM) estimation based on double-cross-correlation processor (DXCP) by an auxiliary function method. SFM estimation is one of the key problems in signal processing on asynchronous microphone arrays. Previously, a DXCP was proposed for obtaining an accurate and robust SFM estimate. The DXCP estimates the SFM by maximizing a cross-correlation (CC) function, where parabolic interpolation is employed to attain the sub-sample time delay (STD) estimate between two observed signals. While, we previously proposed a highly accurate technique of STD estimation based on the auxiliary function method, which reaches a local maximum of a CC function, achieving a better result than the parabolic interpolation. In this paper, we thus extend the DXCP using our approach of STD estimation to improve the performance of SFM estimation. In experiments, we confirm that the proposed method shows a monotonic increase in objective function in the DXCP and achieves better performance than the original DXCP.