Respiration pattern recognition using dual tri-axis accelerometers

Abstract

In this study, respiration pattern is extracted using chest movements. Two accelerometers are used to track chest movements. Accelerometers are placed on the right side of the chest and the dorsal as mirror symmetrical not to be affected by heart beats. The data read from the accelerometer with the microprocessor is transferred to the MATLAB software by wireless communication. Bluetooth is used as a wireless communication method. Respiration pattern is extracted from the data by applying digital filter and axis fusion. Third order low pass Butterworth filter with 0.5 Hz cutoff frequency is applied to accelerometer axes data to eliminate noise. Respiration rate is calculated using filtered data. Results are compared with spirometer which is the golden standard for flow and volume measurements. Correlation coefficient, SKLD( Symmetric Kullback-Leibler Distance) and mean delay values are calculated besides comparing the graphical representations. Validation and comparison tests are applied with two scenarios which are tests on non-moving body and tests on moving body. According to results, 260 ms mean delay, 0.8280 mean correlation coefficient per respiration cycle and 0.7786 correlation coefficient per non-moving tests are calculated when results are compared with spirometer. In the second part of the study, respiration pattern has been tried to be extracted in moving body. The results are compared with spirometer. Tests are repeated with rotational and reciprocating movements. 583 ms mean delay, 0.6845 mean correlation coefficient per respiration cycle and 0.5861 correlation coefficient for per tests are calculated in these tests.