| Summary: | 碩士 === 國立中正大學 === 電機工程研究所 === 107 === This research proposes a sound-based detection to replace the method of traditional "Respiratory training". Compared with the traditional Triflow, this research intends to improve the inconvenience of carrying traditional Triflow and Hygiene issue that the mouth often touches the suction tube by detecting respiratory sounds. Therefore, how to use the respiratory sound to reproduce the same effect as the Triflow is an important topic of this research.
This research attempts to use MFCC and Perceptron to do machine learning to subdivide the respiratory sound into inspiratory, expiratory, and the respiratory interval. The experimental results show that after taking the respiratory signal characteristics by MFCC, the Perceptron model can be used to make Inference. In order to reproduce the realistic respiratory training effect, this paper proposes the calibration of the inspiratory capacity with sound energy further, and calibrates the difference of the inspiratory capacity corresponding to each person’s inspiratory sound to achieve the nearest to the effect of the Triflow. Finally, we can propose a respiratory training system that can replace the traditional Triflow.
|
|---|
