A Study of Vision Based Human Motion Recognition and Analysis

1. Introduction

Vision based human motion recognition is a systematic approach to understand and analyse the movement of people in camera captured content. It comprises of fields such as Biomechanics, Machine Vision, Image Processing, Artificial Intelligence and Pattern Recognition. It is an interdisciplinary challenging field having grand applications with social, commercial, and educational benefits. A wide spectrum of applications demands human motion recognition. The applications are spread over domains like sports, medical, surveillance, content based video storage and retrieval, man-machine interfaces, video conferencing, art and entertainment, and robotics (Aggarwal & Nandkumar, 1988; Aggarwal & Cai, 1999; Aggarwal & Ryoo, 2011; Gavrila, 1999; Poppe, 2007; Turaga et.al, 2008). Some of the applications for highlighting the potential impact of human motion recognition are discussed here.

• •

  Smart Surveillance: In today's surveillance systems, video contents are viewed continuously by human operators. With the increasing number of cameras, it is impossible for humans to monitor all the contents 24 X 7. Generally, the contents are viewed after a mishap to analyse the event. So, there is an intense requirement of smart surveillance systems from the security agencies. Smart surveillance systems can analyse an event online and provide appropriate intimation using computer based human motion and behavioural analysis. Smart surveillance is required for access control in special areas like military territory, distant human identification, counting the persons and congestion analysis, detection of abnormal behaviour at shopping malls, railway stations, hospitals, government buildings, commercial premises, and schools (Makris & Ellis, 2005; Morris & Trivedi, 2008). Nowadays smart home concept is gaining attention of computer vision community to improve the quality of life of the inhabitant (Guesgen & Marsland, 2016).

• •

  Behavioural Biometrics: Nowadays, the use of the gait pattern as a biometric has become popular. The main reason is that the recognition of the gait pattern does not require subject cooperation as compared to the other biometrics (Sarkar et. al 2005).

• •

  Gesture and Posture Recognition and Analysis: For a more advanced natural interface with computers and computerized systems, human gesture and posture recognition is an important key. It has promising applications such as gaming, sign language recognition, controlling devices, and others (Ronchetti & Avancini, 2011; Seperi et. al, 2006).

• •

  Robotics: Human motion analysis plays an important role in robotics for humanoid robot control, to imitate human motions in a robot in virtual and augmented environments (Hoffman, 2010).

• •

  Medical: The medical field uses human motion recognition for the study and analysis of Orthopaedics, Neurology, Musculoskeletal disorders, body posture, and fitness. It is also useful to design intelligent systems to assist elderly people and physically / mentally disabled ones (Najafi et. al, 2003; Lin & Kulic, 2013).

• •

  Sports and Exercise: In sports, motion recognition is useful to analyze athletic movements and to design affordable and efficient frameworks for training (Bertini, 2003). An environment for rehabilitation exercise with a feedback system at remote places or in the presence of an expert is designed (Watanabe 2015). Dao (2016) proposed a monitoring system for the exercises of elderly people. These kinds of systems will definitely be useful for patients and old age people.

• •

  Art and Entertainment: Motion recognition is useful in analyzing, learning, and an emotional understanding of artistic dance movements as in dances like Bharatnatyam, and Salsa. Kale and Patil (2015) have recognized Bharatnatyam dance sequence from depth data. This also helps to increase the effectiveness of a scene, and the alteration of movements required for quality and the impact of acting.