Semantic Communication for Critical 6G Applications Based on Compressed Sensing
AbstractIn numerous mission-critical applications anticipated by future sixth-generation (6G), such as autonomous vehicles, achieving high accuracy in image recognition is essential. Concurrently, minimizing the data traffic, or equivalently, effective data compression due to constraints posed by transmission delay are crucial. To overcome these challenges, this paper proposes a task-oriented semantic communication system dedicated to image data, designed to extract and transmit the information required by the receiver. The goal of the system is to recognize the semantic boundaries of the images. For this For this relevant, a novel semantic encoder, based on compressed sensing (CS) is developed at the transmitter to extract semantic information. Additionally, a novel semantic decoder is proposed in the receiver, utilizing sparse reconstruction techniques to reconstruct semantic information. In contrast to prior studies that focused on conveying a broad spectrum of semantic information related to images along with all extracted features, this approach concentrates solely on isolating the semantic features relevant to the specific target edges. This method generates a sparse feature map, allowing for a reduction in compression rates by a new compressed sensing techniques implementing new sensing matrix based on polar code (SMPC)compressed sensing techniques. Furthermore, the study examines two distinct typical scenarios: noiseless measurements and noisy measurements. Our simulations show that with a lower compression rate, the classification accuracy, and exact recovery probability of 100 % can be attained in both scenarios.