Optimized design of windowed-sinc anti-aliasing filters for phase-preserving decimation of hydrophone data

Passive acoustic monitoring generates large data sets for which decimation is beneficial to analysis and portability for data sharing. Among the goals for effective decimation are avoidance of aliasing in the passband, accurate and complete control of the attenuation profile, phase preservation, and high efficiency in processing. We present an approach to decimator design that addresses each of these goals, and we demonstrate its application to ocean audio recordings. Anti-aliasing is achieved by windowed-sinc filters that also preserve phase. Control of the passband attenuation profile is based on the specification of the maximum allowed attenuation at a certain percentage of the final output Nyquist frequency. The window type is selected to meet the stopband attenuation requirement. Efficiency is achieved through optimization of the anti-aliasing filters applied in each decimation stage, and through parallelization of processing. The best combination of the sinc function's cutoff frequency and the mainlobe bandwidth of the window function generates the shortest qualifying filter, optimizing the trade-off between filter performance and computational load. Parallelization is enabled by applying the overlap-add method to contiguous segments of audio data, consistent with the commonly used storage of contiguous audio data in files of limited duration. Beyond addressing common goals for effective decimation of audio data, the approach presented is deployable in open-source environments.