Research On Applications Of Compressed Sensing In Wireless Audio Sensor Networks Data Acquisition

Wireless audio sensor networks (WASN) is a simple form ofwireless multimedia sensor networks(WMSN), which is responsible forcollecting audio signal in examination area, and extracting the featureparameters according to user needs or transmitting complete audio signal tothe client through the wireless network. It is different from the traditionalwireless sensor network which only needs to collect and transmit a smallamount of scalar data，such as temperature, humidity, and light intensity,WASN faces new challenges brought by high speed audio signal acquisitionand large flow sampling data transmitting. The main challenge is the audiosensor nodes which have limited computing, storage and power resources canhardly meet the requirements of high-speed sampling, real-time processingand transmission of audio signal.In fact the root cause of the above problem lies in the signal samplingmust follow Nyquist principle, which says that sampling frequency can’t lessthan2times the bandwidth of signal. Compressed sensing theory overturnsthe Nyquist sampling theorem, it based on signal sparse property proposes anew framework of signal acquisition and processing. Compressed sensingtheory points out that: a signal which has sparse property can be sampled by afrequency far below the Nyquist frequency, and can be accuratelyreconstructed by the low speed sampled data.Compressed sensing theory opens up a low cost way of signal samplingand processing, incorporating sampling and compression process.Compressed sensing is applied to wireless audio sensor network in this paper,which explores resources saving data collection schemes.The main contents of this paper are listed as follows:(1) Based on compressed sensing theory, a audio signal acquisitionscheme as the core of the data compression is proposed in this paper. In thisscheme audio sensor nodes should sample audio signal with Nyquist frequency, and then achieve high ratios compression of audio signal throughthe compressed sensing linear measurement process. This paper analyzes thesparse property of audio signal in the commonly used sparse basises, designsthe compression method based on Bernoulli binary matrix, build thereconstruction model of noise signal, and simulates the performance ofproposed Bernoulli binary matrix.(2) In order to further reduce the amount of audio data, and reduce thenode energy consumption brought by data acquisition and data transmission,a audio signal acquisition scheme as the core of random sampling areproposed in this paper. In this scheme audio sensor nodes directly randomsample the analog audio signal by a low speed, which is far below theNyquist sampling frequency, and then transmit the random ly sampledsamples to the client. The complex signal reconstruction and signal analysiswould be done by the computer of client. On the basis of previous work, anadditive random sampling time sequence generation method is proposed inthis paper, the equivalent random sampling observation matrix is deduced,and the performance of the proposed measurement matrix are analyzed in thesimulation.(3) In order to verify the feasibility of the proposed data acquisitionschemes, data acquisition experiments were done based on the existing audiowireless sensor network hardware platform, and a PC software was developedusing LABVIEW2011.So a complete audio data acquisition system wereconstructed.Experiment results show that the proposed two audio signal acquisitionschemes are reasonable and feasible, both can realize the multipointdistribution acquisition, real-time transmission and high precisionreconstruction of audio signal. Compared with the scheme of datacompression, the random sampling scheme is more suitable for the audiowireless sensor networks, which with limited resources.