Sub-Nyquist non-uniform sampling for low-cost sound monitoring

Low-cost sound monitoring is required for maintaining machinery. We aim to reduce the cost of monitoring by reducing the sampling rate, i.e., sub-Nyquist sampling. Monitoring based on sub-Nyquist sampling requires two sub-systems; a sub-system on-site for sampling machinery sound at a low rate, and a sub-system off-site for reconstructing the original sound from the sub-sampled signal. In this paper, we focus on the former and clarify what sampling methods are effective. In areas such as communication-signal processing, it has been clarified that co- prime sampling and sparse ruler sampling exhibit high spectral- reconstruction performance for wide-sense stationary signals. However, it is not known whether these methods are suitable for signal reconstruction in sound monitoring. In this study, these sampling methods were applied to machinery sounds, and their effectiveness was clarified. The experimental results indicate that non-uniform sampling is effective for sound monitoring and that the reconstruction performances of co-prime sampling and sparse ruler sampling are on the same level as that of random sampling. By taking into consideration that it is easy to implement co-prime sampling and sparse ruler sampling on analog-digital-converters, it can be concluded that these two methods are the best choices.