A Low Cost Anthropometric Body Scanning System Using Depth Cameras

© 2018 IEEE. The measurement of the human body and categorisation of body types has become a key indicator of health risks over the past century. At first, being of interest to a mathematician, the body shape and anthropometrics was defined in order to create 'the average man', an idea of what the best proportions of the typical male should be composed of during the 18th century. This would later evolve into what is commonly known as Body Mass Index (BMI) following the interests of life insurance companies trying to define a risk of fatality. The development of body type categorisation known as BMI was soon thereafter created. BMI has been a standard of measurement in a range of applications from health and well-being industries to government policy and national health surveys. However more recently, BMI has come under question of its reliability due to the number of parameters used to classify an individuals body type. This has been found in some cases to either overestimate or underestimate the category in which an individuals body mass index is defined. In this chapter, a simple non-invasive measurement tool capable of estimating subject specific body parameters is used to increase the accuracy of BMI measurement. The formulation of synthetic data sets is used to create a data base of population characteristics models. This database was then fed into machine learning models to perform regression of anthropometric measures from images acquired from depth sensors such as Kinect.