Research on the Application of Integrated Storage and Computing Chip Based on Image Recognition

This paper proposes a chip deployment based on a regularized optimization algorithm for image recognition. This algorithm solves the problem of difficult deployment of storage-computing integrated chips. The information entropy of the weight data of the regularization algorithm can be increased to make full use of the analog characteristics of the memristor unit. The system realizes the concrete circuit on the programmable device by using the hardware description language and adopting the top-down design method. Introducing this method to the keyword recognition task in the programmable linear memristor-based storage-computing integrated system-on-chip finally achieved a recognition accuracy of about 97%. Improve the recognition accuracy by about 4%.