Investigation of sampling frequency requirements for acoustic source localisation using wireless sensor networks

There is considerable interest in the use of wireless sensor networks (WSNs) for distributed sound capture and acoustic source localisation (ASL) where array elements are spaced over a large area. High sampling rates, such as digital audio at 44.1 kHz, pose a major challenge for efficient wireless personal area network (WPAN) standards such as IEEE 802.15.4 (Zigbee) with an absolute maximum data throughput of 250 kbps. This paper investigates the effect of sampling frequency on the accuracy of time delay estimation using different algorithms in the time domain, such as basic cross correlation (BCC) and generalised cross correlation (GCC), frequency and content based features such as envelope, including generalised phase spectrum (GPS) and envelope-GPS (EGPS). Experimental and simulation studies have been undertaken which show that frequency domain and content based features algorithms can achieve more accurate time delay estimation at low sampling frequencies than time domain algorithms if the appropriate signal contents are extracted. Therefore they are more appropriate for wireless ASL applications.