Feature Diversity for Optimized Human Micro-Doppler Classification Using Multistatic Radar

Abstract

Feature diversity for optimized human micro-doppler classification using multistatic radar. IEEE Transactions on Aerospace and Electronic Systems, 53(2), pp. 640-654. (doi:10.1109/TAES.2017.2651678) This is the author's final accepted version.There may be differences between this version and the published version. FEATURE DIVERSITY FOR OPTIMIZED HUMAN MICRO-DOPPLER CLASSIFICATION USING MULTISTATIC RADARFrancesco Fioranelli (1) , Matthew Ritchie (2) , Sevgi Zübeyde Gürbüz (3) , Hugh Griffiths (2) ( AbstractThis paper investigates the selection of different combinations of features at different multistatic radar nodes, depending on scenario parameters, such as aspect angle to the target and signal-to-noise ratio, and radar parameters, such as dwell time, polarisation, and frequency band. Two sets of experimental data collected with the multistatic radar system NetRAD are analysed for two separate problems, namely the classification of unarmed vs potentially armed multiple personnel, and the personnel recognition of individuals based on walking gait. The results show that the overall classification accuracy can be significantly improved by taking into account feature diversity at each radar node depending on the environmental parameters and target behaviour, in comparison with the conventional approach of selecting the same features for all nodes.

Feature diversity for fall detection and human indoor activities classification using radar systems

“…the centre of mass of the spectrograms and the intensity of the signatures around it. The mean and the standard deviation of these two quantities have been previously used for human micro-Doppler classification [15,17].  Entropy of the spectrogram image and skewness of the histogram containing the intensity samples.…”

“…Seven different subjects took part to this experiment, aged between 23 and 40 years old. It is known that micro-Doppler signatures can change significantly for the same action at different aspect angles [15], especially at aspect angles that approach 90° to the radar line of sight. Multistatic radar has been suggested as possible solution to approach this issue, where different radar nodes collect simultaneous signatures of the subject from different aspect angles, as well as different classes of sensors that can be more tolerant of the aspect angle issue.…”

Bistatic human micro-Doppler signatures for classification of indoor activities

“…These have been shown to be effective for different classification problems such as personnel recognition [18] and unarmed vs armed personnel classification [19], and have the advantage of not requiring any pre-processing or parameter tuning in the feature extraction algorithm. The Doppler centroid estimates the center of gravity of the micro-Doppler signature, and the Doppler bandwidth calculates the energy extent of the signature around the centroid [17]. These parameters were calculated as in equations (1) and (2), where S(i,j) represents the value of the spectrogram at the i th Doppler bin and j th time bin and f(i) is the value of the Doppler frequency at the i th bin.…”

“…Three different classifiers were used to process the feature samples, namely Binary Tree (BT), Naïve Bayes (NB), and Nearest Neighbor with 3 neighbors (3NN). More details on the implementation of these classifiers can be found in [17,20]. The classifiers were trained with 70% of the available data and tested with the remaining 30%, this process was repeated using 100 monte-carlo simulations that used different, randomly selected subset of samples for training and testing.…”

“…Furthermore, it is shown how different subsets of features at different nodes (monostatic and bistatic) are required to provide maximum accuracy. This opens up wide scope for further work in exploring this 'feature diversity' [17] in human indoor activities classification scenarios, aiming to characterize the effect on the overall accuracy of the many operational parameters.…”