Resilient image sensor networks in lossy channels using compressed sensing

Abstract

Data loss in wireless communications greatly affects the reconstruction quality of wirelessly transmitted images. Conventionally, channel coding is performed at the encoder to enhance recovery of the image by adding known redundancy. While channel coding is effective, it can be very computationally expensive. For this reason, a new mechanism of handling data losses in wireless multimedia sensor networks (WMSN) using compressed sensing (CS) is introduced in this paper. This system uses compressed sensing to detect and compensate for data loss within a wireless network. A combination of oversampling and an adaptive parity (AP) scheme are used to determine which CS samples contain bit errors, remove these samples and transmit additional samples to maintain a target image quality. A study was done to test the combined use of adaptive parity and compressive oversampling to transmit and correctly recover image data in a lossy channel to maintain Quality of Information (QoI) of the resulting images. It is shown that by using the two components, an image can be correctly recovered even in a channel with very high loss rates of 10%. The AP portion of the system was also tested on a software defined radio testbed. It is shown that by transmitting images using a CS compression scheme with AP error detection, images can be successfully transmitted and received even in channels with very high bit error rates.