DNN-Based Human Face Classification Using 61 GHz FMCW Radar Sensor

• Published in: IEEE sensors journal Vol. 20; no. 20; pp. 12217 - 12224
• Main Authors: Lim, Hae-Seung, Jung, Jaehoon, Lee, Jae-Eun, Park, Hyung-Min, Lee, Seongwook
• Format: Journal Article
• Language: English
• Published: New York IEEE 15.10.2020; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Antennas; Artificial neural networks; Classification; Continuous radiation; Deep neural network; Face; face classification; Machine learning; Millimeter wave radar; Millimeter waves; Radar; Radar antennas; Radar signal processing; Receiving antennas; Sensors; Support vector machines
• ISSN: 1530-437X; 1558-1748
• DOI: 10.1109/JSEN.2020.2999548

In this paper, we propose a method for classifying human faces using a small-sized millimeter wave radar sensor. The radar sensor transmits a frequency-modulated continuous wave signal operating in the 61 GHz band and it receives reflected signals using spatially separated receiving antenna elements. Because the shape and composition of the human face varies from person to person, the reflection characteristics of the radar signal are also distinguished from each other. Therefore, training a deep neural network (DNN) using signals received from multiple antenna elements enables classification of different human faces. With our trained DNN model, eight human faces can be classified with an accuracy of 92%. We also compare the performance of the proposed method with conventional machine learning techniques (e.g., support vector machine, tree-based methods) and confirm that our method has higher classification accuracy.