In this communication, a low-cost radar-sensor-based apparatus for contactless hand gesture recognition via Doppler signature analysis is proposed. The raw reflected signal, after some pre-processing, is analysed via its time-frequency representation, known as spectrogram. This information is then exploited to train a convolutional neural network (CNN) to perform the classification step. The whole procedure was tested on an in-house experimental data set composed of four different hand gestures, showing good performance and reaching an accuracy of approximately 97%. Finally, the classification performance was tested also in a cluttered environment which includes the presence of a strong echo close to the target.
Hand Gesture Recognition via Radar Sensors and Convolutional Neural Networks
Franceschini S.;Ambrosanio M.;Vitale S.;Baselice F.;Gifuni A.;Grassini G.;Pascazio V.
2020-01-01
Abstract
In this communication, a low-cost radar-sensor-based apparatus for contactless hand gesture recognition via Doppler signature analysis is proposed. The raw reflected signal, after some pre-processing, is analysed via its time-frequency representation, known as spectrogram. This information is then exploited to train a convolutional neural network (CNN) to perform the classification step. The whole procedure was tested on an in-house experimental data set composed of four different hand gestures, showing good performance and reaching an accuracy of approximately 97%. Finally, the classification performance was tested also in a cluttered environment which includes the presence of a strong echo close to the target.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.