Scalable compressive sensing-based multi-user detection scheme for Internet-of-Things applications

Abstract

Rapid growth in the number of users (largely sensor nodes) of most Internet-of-Things (IoT) applications is widely expected in recent years. 3G or 4G protocols are designed for regular Human-to-Human communication purposes instead of low-rate, Machine-to-Machine (M2M) communication, which is the basis of IoT. Comparatively, traditional spread-spectrum protocols such as Code-Division Multiple Access (CDMA) are considered applicable for the majority of IoT applications. However, they are unsuitable to be utilized in the expected massive-IoT scenarios. The traditional Multi-User Detection (MUD) approach used in these protocols has poor scalability. When the number of users reaches hundreds, the complexity and hardware cost of traditional MUD become impractically high. In this paper, we employ the concept of Multiple Measurement Vector Compressive Sensing (MMV-CS) to exploit the feature of user sparsity in spread-spectrum-based IoT applications. By reformulating the CS detection model, a memory reduction of approximately 9,400 times can be achieved. Along with other hardware cost-down, the proposed simplified structure of the reconstruction model also allows a faster detection speed (~10x to 1000x improvement) while a satisfactory level of Bit-Error-Rate (BER) is still maintained.