A 22.8GOPS 2.83mW neuro-fuzzy Object Detection Engine for fast multi-object recognition

Abstract

A neuro-fuzzy Object Detection Engine (ODE) is proposed as the pre-processing accelerator of multi-object recognition processor to reduce the computational complexity. It performs a fast and robust neuro-fuzzy object detection algorithm with Motion Estimator (ME) and Visual Attention Engine (VAE) within 1ms. The mixed mode implementation achieves 22.9GOPS 2.83mW ODE, and reduces the area by 59% and power consumption by 44%. The ODE can increase the frame rate by 2.09x and reduce power consumption by 38% of the multi-object recognition processor.