Improvement of body posture changes detection during ambulatory respiratory measurements using impedance pneumography signals

Abstract

Impedance pneumography could be used for measuring respiratory parameters quantitatively in ambulatory conditions. It was noted that body posture affects the calibration coefficient connecting the measured impedance values and their first derivatives with volume and flow reference signals. Standard techniques for automatic detection of body posture and activity usually require additional motion sensors. However, in terms of the measurement comfort, less number of sensors is needed. Single sensor mounted on the chest provides good results, however its accuracy decreases during frequent changes of body posture. The aim of this study was to assess the possibility to detect body posture changes using the impedance signal itself, without any other devices or the active cooperation of the person being studied and prospectively improving the body posture change detection method using single motion sensor (e.g. 3D accelerometer). Fifteen healthy students (11 males) performed two body posture changes - get-ups and stand-ups. Six classification techniques were checked for prediction accuracy. It was found that artificial neural networks provided the best overall accuracy (90%).