Design and Development of a Sitting Posture Recognition System.

Sitting posture recognition can be used to evaluate the awareness of a person carrying out a task, such as working or driving, and can aid in avoiding accidents or other health risks, such as musculoskeletal disorders. In addition, sitting posture can reveal wellness or unhealthiness for the elderly and mobility disabled individuals. This paper focuses on body posture monitoring, by acquiring the pressure distribution of a sitting person with thirteen piezoresistive sensors placed on a seat. The measurements from the sensors passing through a microcontroller unit fed several machine learning techniques in order to discriminate among five sitting postures (upright, leaning left, leaning right, leaning forward and leaning backward). Experiments with body postures from twelve individuals (six men and six women) of different Body Mass Index (underweight, normal and overweight) were conducted. The developed classifiers achieved average discrimination accuracy over 98% among the aforementioned five body postures.