Date of Completion
Compressive sensing, Memristor, Analog Computing, Process variation, Sparsity estimation, Low power hardware design, Video Streaming, Orthogonal matching pursuit, Prior information
Field of Study
Doctor of Philosophy
The amount of sensory signal is increasing dramatically as we’re stepping into the era of Internet of Things (IoT). Compressive Sensing (CS), feature as Sub-Nyquist sampling rate and low complexity sensing architectures, is very promising for these kinds of applications where resources are restricted. Through applying this novel compression technology, data size of sensory signals are largely compressed such that it is very efficient within the signal processing, data transmitting and storage processes. Compared to conventional codec method, CS technique requires less hardware resources and achieve lower power consumption within sensor nodes.
However, there are several bottle-necks of existing compressive sensing implementation, which discourage the utilization in practical applications. Based on our continues studies, memristor devices are exploited to design the compressive sensing system for sensory signals with multiple functions.
First of all, memristor is utilized as random number generator for sensing matrix, and in-memory computing is also achieved based on array structure. A new memristor model is proposed to evaluate its feasibility of being utilized with CS applications under fabrication variations. Second, a comprehensive CS system is proposed for the application of video streaming. Inside the proposed system, memristor devices are also used to implement the control logic for real time compression rate optimizations. Afterwards, a new prior algorithm is proposed by us to further improve the CS process with higher compression ability. The utilization of memristor is extended to the generation of prior information. Evaluation results demonstrate the advantages of our work in different aspects. In general, our proposed CS system can achieve higher energy efficiency, less hardware complexities, and with very good recovery quality, compared to existing implementations of both CS system and conventional codec method.
Qian, Fengyu, "Exploiting Memristors for Compressive Sensing Applications" (2019). Doctoral Dissertations. 2393.