A Robust Abnormal Behavior Detection Method Using Convolutional Neural Network

Abstract

A behavior is considered abnormal when it is seen as unusual under certain contexts. The definition for abnormal behavior varies depending on situations. For example, people running in a field is considered normal but is deemed abnormal if it takes place in a mall. Similarly, loitering in the alleys, fighting or pushing each other in public areas are considered abnormal under specific circumstances. Abnormal behavior detection is crucial due to the increasing crime rate in the society. If an abnormal behavior can be detected earlier, tragedies can be avoided. In recent years, deep learning has been widely applied in the computer vision field and has acquired great success for human detection. In particular, Convolutional Neural Network (CNN) has shown to have achieved state-ofthe-art performance in human detection. In this paper, a CNN-based abnormal behavior detection method is presented. The proposed approach automatically learns the most discriminative characteristics pertaining to human behavior from a large pool of videos containing normal and abnormal behaviors. Since the interpretation for abnormal behavior varies across contexts, extensive experiments have been carried out to assess various conditions and scopes including crowd and single person behavior detection and recognition. The proposed method represents an end-to-end solution to deal with abnormal behavior under different conditions including variations in background, number of subjects (individual, two persons or crowd), and a range of diverse unusual human activities. Experiments on five benchmark datasets validate the performance of the proposed approach.

Sensing Your Touch: Strengthen User Authentication via Touch Dynamic Biometrics

“…Experimental results showed that the idea of using touch dynamics biometrics to support user authentication in a mobile device or application context is feasible. Our future work is to use deep learning techniques [23]- [26] to automatically extract representative features to build more accurate authentication models.…”

Transfer Deep Learning Along With Binary Support Vector Machine for Abnormal Behavior Detection

“…As shown in Fig. 1, four main kinds of deep learning architectures are utilized for abnormalities detection namely: Convolution Neural Network (CNN) [11]- [13], Autoencoder (AE), Generative Neural Network (GAN) [14] and Recurrent Neural Network (RNN) [15]. Training a new deep learning model from scratch demands a considerable amount of data, high computational capabilities, and very long processing time.…”

“…Additionally, Li [50] proposed a deep spatiotemporal architecture, which has convolutional and recurrent nature for understanding pedestrian behavior in a crowd. In [11], authors study abnormal behavior detection in different situations such as various background settings and number of subjects using CNN. Zenati et al [14] proposed a new framework based on Bidirectional GAN, which simultaneously learns an encoder E that maps samples x in latent representation z together with a generator G and a discriminator D during training.…”

Revisiting crowd behaviour analysis through deep learning: Taxonomy, anomaly detection, crowd emotions, datasets, opportunities and prospects

“…In [75] , the action anomaly detection problem is solved using a hybrid anomaly detection and behaviour classification approach. A CNN was trained to distinguish among six different classes of anomalous behaviours, given some frames of an anomalous video sequence.…”