Resource-saving audio signal acquisition methods based on compressed sensing theory

Abstract: In order to reduce the cost by multimedia signal sampling, processing, transmission of computing, storage, electricity, bandwidth resources in an environment-monitoring system built on a wireless sensor networks, this paper analyzed the limitations of the existing signal processing theory and random sampling methods. An improved additive random sampling method and the sampling time sequence obtained by this method had the same probability density function with the existing random sampling method. The proposed method effectively avoiding the phenomenon of adjacent sampling time interval was too large or too small, and each sequential sampling time had a clear causal relationship. On this basis, a novel resource-saving audio signal acquisition method was designed. In the proposed signal acquisition method, the audio signal acquisition was completed by cooperation between the sensor nodes and the sink nodes. The sensor nodes took the low frequency random sampling with the sparse audio signal, and the sink nodes reconstructed the original signal with high probability by using the received random sampling data. Then a test system was established with 13 acquisition nodes and a sink node based on the Zigbee network technology, and this system was used to implement the remote, wireless, and distributed acquisition of the crawl acoustic signal of Tribolium castaneum Herbst adults in grain barrel. In the test system, the proposed signal acquisition method was compared with the existing data compression method. Results showed that if the maximum data reconstruction error was less than 0.5%, the packet loss rate was less than 10% and each acquisition node only sampled one acoustic signal, then the packet loss rate was 9.1%and the maximum reconstruction error was 0.44% by using the proposed signal acquisition method. However, if using the existing data compression method, packet loss rate and the maximum reconstruction error were 9.3% and 0.46% respectively. Two acquisition methods had similar performance, and could achieve the remote, wireless, real-time acquisition for 13 acoustic signals. For the proposed audio signal acquisition method, when the sampling frequency was as low as 196 Hz, the packet loss rate and the maximum reconstruction error were 21.6% and 0.48%, respectively. But for the proposed signal acquisition method, the sampling frequency was only 586Hz, and it effectively reduced the resource consumption of acoustic signal acquisition. The method proposed in this paper can provide references especially for the wireless sensor networks with limited resources.