A 256 pixel, 21.6 μW infrared gesture recognition processor for smart devices

• Main Authors: Ba, Ngoc Le, Oh, Sechang, Sylvester, Dennis, Kim, Tony Tae-Hyoung
• Other Authors: School of Electrical and Electronic Engineering
• Format: Article
• Language: English
• Published: 2020
• Subjects: Engineering::Electrical and electronic engineering; Gesture Recognition; Passive Infrared
• Online Access: https://hdl.handle.net/10356/143356
• Institution: Nanyang Technological University

This paper presents a low-power digital signal processor for gesture detection using infrared (IR) sensors. IR compatible detecting algorithms are introduced to reduce power consumption and provide accurate detections for various types of gestures including sweeping in 8 directions, zooming in/out and a unique wake-up gesture. To further reduce the power, the proposed algorithm is capable of setting the processor into idle mode when there is no motion and resume active mode upon wake-up request. Test chips were fabricated in a 65 nm commercial TSMC process and occupy an area of 580 μm × 580 μm. Measurement results taking input from a 16 × 16 Heiman IR sensor array show that the proposed gesture recognition processor is able to detect correct gestures at 100 MHz maximum clock frequency. Under normal working condition, the processor consumes 21.6 μW in active mode and 7.2 μW in idle mode while processing 8-bit 16 × 16 input data array at 30 fps.