An Ultra-Low-Power Design of Smart Wearable Stereo Camera

This paper presents the design of an Ultra-Low-Power Smart Wearable Multisensory Stereo Camera system, which features the capabilities to be AI-enabled. In this system, two 5MP Omni vision cameras are connected to the same DCMI interface of a dual-core cortex ARM STM32 microcontroller. The M7 processor can be configured to capture images from both cameras in a time-synchronized manner, run a Deep neural network, and perform onboard image recognition and object detection. The M4 processor can be interfaced with other sensors such as a motion detector, 6D IMU with a machine learning core, a BLE 5.0 module to communicate with the user, etc. The onboard 6D IMU contains a machine learning core to host a decision tree classifier and relieve some processing load on the M4 processor. The design constraints, the component selection, and interfacing details are discussed in this paper with the complete schematic and layout files so that this stereo camera can be reproduced by the researchers of the common domain and contribute to the development of an open-source blind navigation system.