Chih-Chieh Huang, Wen-Yen Lin and Ming-Yih Lee
In this work, an accelerometer-based respiratory detection algorithm was developed when the subjects are in both static postures and dynamic activities. The algorithm was also veri-fied with a wearable smart clothes on 30 testing cases from male and female subjects in static postures and 5 testing cases when subjects were in dynamic activities, such as walking and running. More than 97% of accuracy is achieved by compar-ing with the detection results from respiratory belt conducted simultaneously in static cases and more than 85% of accuracy is achieved in dynamic cases. This clothes and respiratory detection is extremely useful to monitor human’s respiration when the subjects are in sleep and hence could be used for the detection of sleep apnea syndrome. Moreover, it could also help to monitor the respiration frequency of the subjects in their activities of daily life (ADLs).
In recent years, tracking respiratory signals has gradually attracted attention in the application of medical care. A spi-rometer is the most commonly used instrument in lung func-tion test to measure air-flow velocity and the volume of lung clinically. However, there is no easy and comfortable way to obtain respiratory signal continuously. In the diagnosis of obstructive sleep apnea syndrome (OSAS), measurement of nasal-oral air flow along with respiratory belt detection are used in polysomnography (PSG) for the continuous monitor-ing of breathing signals and which are not comfortable at all for long-term monitoring especially in the sleep. There are other technologies for respiratory detection, such as detection of slight chest movement through radar, vision-based detec-tion and estimated from ECG signal, but the usages of these technologies either are restricted on the sites with proper instruments installation required or are lack of volume infor-mation for some applications. Consequently, they are not easy to be implemented as a wearable device for easy and comfort-able usage.
When people are breathing, the diaphragm moves up and down accordingly. Diaphragm performs an important func-tion in respiration: as the diaphragm contracts, the volume of the thoracic cavity increases and air is drawn into the lungs, that is the inhalation; as the diaphragm relaxes, it causes the tissue to put pressure on the lungs to expel the air, and that is the exhalation. As the results, the respiratory signal can be monitored. The works on accelerometer-based respiratory signal detection to detect the movement of the diaphragm were not that much. Some research works used accelerometer to measure respiratory frequency during speech or installed accelerometer on the seat-belt of the driver seat in the car to detect the respiratory signal and further process with Empiri-cal Mode Decomposition (EMD) algorithm to monitor driver's respiration.
In this study, wearable smart clothes instrumented with an accelerometer sensor is used to verify the developed respirato-ry detection algorithm. The smart clothes is easy to use and comfortable to wear and hence it is suitable for long-term monitoring. The validations was conducted with healthy sub-jects in the supine position to emulate the sleep posture. As a matter of fact, the system would be very useful for respiration monitoring during sleep and which could be applied for many sleeping related issues, such as sleep disorder problem, sleep quality monitoring, and especially for the OSAS.
The respiratory detection in dynamic activities, such as walking and running, were also developed. The Ensemble Empirical Mode Decomposition (EEMD) algorithm were used to dissemble the measured signals into the intrinsic mode functions (IMF), which consists of different components of signals in different frequency. Comparing with the data de-tected from respiratory belt, the accurate signals containing the actual respiratory rate can be found in one of the IMF signals. However, the most challenged part is to decide which IMF signal to choose for accurate detection. Therefore, more testing were conducted in dynamic cases where we had the testing subjects walked or ran on the treadmill on different speed. After the tests, we can concluded that when the speed are greater than 4 km/hr, the IMF7 signal can deliver more accurate information of respiratory rate. As the speed is below 4km/hr, IMF7+8 could provide more accurate signal. In all cases (speeds), more than 85% of accuracy can be achieved when the right IMF components are selected for the respirato-ry signal detection.