Autonomous Human and Animal Classification Using Synthetic 2D Tensor Data Based on Dual-Receiver mmWave Radar System

Abstract

Human and animal classification under the rubble is necessary to successfully rescue survivors post-disaster, and interference from other animals is becoming an issue in noncontact monitoring with radar. Many animals in indoor and outdoor environments have characteristics similar to humans, where they are easily mistaken for human targets, which would trigger a false alarm. A novel human and animal classification through single-receiver and dual-receiver mmWave radar at 77 GHz is presented. The system uses feedback signal responses from targets with dual-receiver mmWave radar and classifies human and animal features using a convolution neural network (CNN) based on synthetic 2D Tensor data. The performance is compared to the classification result using a deep learning and backpropagation neural network (BPNN) from a single mmWave radar dataset under the measurement distance and number of objects. Our experimental results showed that using dual receivers was very useful and performed well in classifying humans and animals, with an average accuracy rate of 99% in the classification of 2 and 3 classes, 83% in 6 classes, and 68% in 10 classes. The performance showed an excellent result considering that this system’s application avoids false alarms in human rescue applications during post-disaster periods and exceeds the accuracy of several previous studies.

Keywords

• Deep learning
• human and animal classification
• mmWave radar
• synthetic 2D Tensor
• dual-receiver