Abstract
Human action or activity or behavior analysis, recognition, and understanding are very important research areas in the field of computer vison, internet of things (IoT) sensor-based analysis, human-computer interaction (HCI), affective computing, intelligent system, healthcare facilities, and so on. There is much importance in human action recognition. This chapter introduces the core aspects of human action or activity recognition (HAR). These are split into two different domains: computer vision-based action/activity recognition, action localization, etc.; and wearable IoT sensor-based HAR. Though cameras are also sensors that provide vision-based information, the author puts camera-based or vision-based methods in another category. The chapter introduces core challenges and strategies to move forward.
TopIntroduction
Human action or activity or behavior analysis, recognition, understanding are very important research areas in the field of computer vison, Internet of Things (IoT) sensor-based analysis, human-computer interaction (HCI), affective computing, intelligent system, healthcare facilities, and so on (Ahad et al, 2012; Ahad et al, 2011; Ahad et al, 2019). There are lots of importance on human action recognition. For example,
- •
Video surveillance in various facilities
- •
Games and interactive systems
- •
Hospital and medical applications
- •
Smart-home, rehabilitation centers
- •
Assisted living
- •
Context-awareness
- •
Fall down detection and recognition
- •
Activity of Daily Living (ADL)
- •
Entertainment fields as movie, animation, 3D TV
- •
Intelligent Transport Systems (ITS)
- •
Robotics and automation
- •
Abnormal behavior analysis
- •
Crowd scene analysis
- •
Affective computing
- •
Emotion analysis
- •
Pain analysis
- •
Gait recognition
- •
Gait understanding for healthcare applications
- •
Man-machine interaction
- •
Gesture analysis
- •
Scene analysis
- •
Egocentric applications
- •
Hand movement understanding
- •
Sign language recognition
TopSensors For Human Activity Recognition
In this section, we split the human action or activity recognition approaches into two broad areas, namely – vision-based domain (based on video cameras) and wearable / IoT sensor-based domain (based on wearable sensors having accelerometer, gyroscope, or other sensors). Methodologies and sensors are widely varied in both arenas. There are only a few approaches where these two modalities are blended.