April 27, 2012

Verification of Simulated Acoustic Environments Utilizing Cross-Correlation and Power Spectral Density

Jonathan Soli, Advisor: Kevin Donohue

This project compares experimental and simulated audio array signals using quantitative and qualitative analyses. Simulator performance metrics included microphone propagation delays, secondary echoes, and frequency-dependent attenuation. Experimental and simulated recordings of transients and white noise were acquired with several microphone (mic) distributions in reverberant and non reverberant environments. Each simulated and experimental mic signal was cross-correlated with the signal of the mic closest to the source to obtain propagation delays and secondary echo locations. Paired T-tests verified that the simulated delays and secondary echo locations were within a 95% confidence interval. Power Spectral Density (PSD) plots were generated for each recorded and simulated signal to evaluate frequency-dependent attenuation. Results showed that the simulated PSD shapes mimicked low frequencies well, but not above 2 kHz. High-frequency roll-off was evaluated in non-reverberant trials by applying a leastsquares regression line to frequencies of the PSDs between 2 kHz and 20 kHz. The resulting simulated and experimental slopes were compared with T-tests and verified to be within 95% confidence. Overall, the simulator emulated propagation delays, secondary echoes and frequency-dependent attenuation well. However, improvements could be made by including audio cage structure and frequency-dependent absorption of the walls in the simulated model.