Dynamic synchronous averaging for enhancement of periodic signal under sampling frequency variation

In this paper, we present a novel method of estimating a room impulse response (RIR) in noisy environments by playing a known periodic signal and recording it for a long time. As is well known, a periodic signal can be easily enhanced, even in a noisy environment, by synchronous averaging. However, in a long-time recording, the sampling frequency of the recording device might fluctuate temporally, which leads to a synchronization error in averaging and degrades the performance of enhancement. To solve this problem, we estimate the time shift between the played-back periodic waveform and the observed signal by their cross-correlation, period by period, and apply synchronous averaging while compensating for the shift dynamically. We also introduce an iterative approach in dynamic synchronous averaging to further improve the performance. In simulation experiments, we confirm that the proposed method effectively enhances the signal and contributes to RIR estimation with high accuracy.