ABSTRACT
Understanding human activities and behavior has long been a research goal. Employing a person to monitor other persons’ activity 24 h a day is unrealistic; therefore, the automated, automatic recognition of human activities is important and necessary. Although it is natural for humans to recognize their activities, it is not an easy task for a computer. Computers need to analyze the information gathered from sensors and infer the ongoing activity. Sensors’ noises and variances make activity recognition (AR) even more complex for computers. According to the sensing techniques, the existing AR can be roughly divided into three categories: video sensor–based activity recognition (VSAR), wearable sensor–based activity recognition (WSAR), and object usage–based activity recognition (OUAR). The chapter discusses the strengths and limitations of VSAR, WSAR, and OUAR. Based on our comparisons, fusing these three approaches seems to be the most promising solution for complex AR applications.
