Path-based channel estimation for acoustic OFDM systems: Real data analysis

We address detection of acoustic OFDM signals using a channel estimation method based on a physical model of multipath propagation rather than an equivalent sample-spaced model. The path identification (PI) algorithm focuses on explicit estimation of delays and complex amplitudes of the channel paths. We apply this algorithm, along with the conventional least squares (LS) and orthogonal matching pursuit (OMP) to a set of signals recorded over a mobile acoustic channel. We demonstrate excellent performance of the PI algorithm and show that its complexity is considerably lower than that of the OMP algorithm. The PI algorithm consistently outperforms the conventional LS and compares favorably with the OMP algorithm in terms of the mean-squared data detection error observed for a varying number of OFDM carriers and receiver array configurations.