Sparsity Aware Dynamic Gesture Classification Using Dual-band Radar

Recognition of Dynamic Hand Gesture Based on Mm-Wave Fmcw Radar Micro-Doppler Signatures

“…The selection of classification algorithm mainly depends on the types of the extracted features from hand gesture signals. In traditional two-phase classification methods, empirical features [16], the principal component analysis features [17], and the sparse features [18] are first extracted and then fed into a classifier, such as support vector machine (SVM), random forest or decision tree to accomplish dynamic HGR. The emerging deep learning algorithm, such as convolutional neural networks (CNN), has brought a breakthrough in various fields [19], which is regarded as an effective method for dynamic HGR.…”

Dynamic Hand Gesture Classification Based on Multistatic Radar Micro-Doppler Signatures Using Convolutional Neural Network

“…Numerous noncontact approaches based on radar have been proposed [1], which cause less distraction to the user and provide a more comfortable experience than approaches using wearable sensors. Most algorithms of dynamic hand gesture recognition with radar sensors are based on micro-Doppler analysis [2][3][4][5][6]. In the conventional two-phase classification algorithms, features such as the empirical features [2], the principal component analysis based features [3], and the sparse features [4] [5] are first extracted from the time-frequency domain and then fed into an off-the-shelf classifier, such as the nearest neighbor, the support vector machine, and the decision trees.…”

“…Most algorithms of dynamic hand gesture recognition with radar sensors are based on micro-Doppler analysis [2][3][4][5][6]. In the conventional two-phase classification algorithms, features such as the empirical features [2], the principal component analysis based features [3], and the sparse features [4] [5] are first extracted from the time-frequency domain and then fed into an off-the-shelf classifier, such as the nearest neighbor, the support vector machine, and the decision trees. The emerging deep neural network, including CNNs, which have enjoyed great successes in various fields [7], is regarded as another powerful tool for dynamic hand gesture classification.…”

“…In recent studies of human activity classification [5] [9][10][11], multistatic radar and multi-angle radar systems have shown their potential to provide higher classification accuracy than single monostatic radar sensors, thanks to the diversity of viewing angles. The key to high-accuracy classification with a multistatic radar is a fusion strategy that makes full use of the signals from different receivers.…”

“…In [9], empirical features extracted from three perpendicular radar sensors are concatenated and fed into a decision tree for dynamic hand gesture classification. Similar approach is adopted in [5] where the sparse features of hand gestures from multi-angle radar receivers are concatenated before being sent to the nearest neighbor classifier. While the aforementioned studies apply fusion to the features (mid fusion), late fusion is adopted in [10] [11], where the decisions, i.e., the classification results obtained from different receivers, are fused to give the final output.…”