Perception,Recognition And Application Of Non-Speech Body Sounds

We call the internal voice of the body as Non-Speech Body Sounds,including breathing,eating sound,drinking sound,cough,sleep and a series of subtle sounds in human body in daily life.Some of these internal sounds are often able to transfer a lot of signals about the health.So,Non-Speech Body Sounds research is currently attracted to the academic and industry attention.In academia,more and more research firms are beginning to focus on the use of wearable devices to collect Non-Speech Body Sounds for human-related status and behavioral research,and attempt to use this study to a certain extent to find the human physiological or psychological biological rhythm and hidden disease.In the industry,a variety of small and sensitive sensors give more possibilities for our research,the use of these sensors can be accurate and convenient to find a variety of human internal voice,which provides the basis for our research.The current results have been made using these sensors for the perception of Non-Speech Body Sounds,but this perception is not satisfactory.In addition,under such perceptual conditions,the existing work is identified by the method of machine learning,and the result is not accurate.Based on this research,this thesis aims to sense and identify Non-Speech Body Sounds,improve the efficiency of perception through homemade wearable devices,and improve the recognition by time domain,frequency domain and Melody cepstral coefficients analysis.This thesis uses the self-made wearable necklace to sense the Non-Speech Body Sounds,which uses a small device and a suitable microphone for the perception of the Non-Speech Body Sounds to improve the perceived effect.Through the data pro-cessing analysis of the unique technical framework,the perceived Non-Speech BodySounds are analyzed in the time domain,the frequency domain and the Melody cep-strum coefficients,and then the feature extraction and selection are carried out using the frame dimension and the window dimension.This thesis first analyzes the specific research problems involved in Non-Speech Body Sounds,summarizes and analyzes the current research situation.On this basis,this thesis puts forward the basic idea:the motivation of Non-Speech Body Sounds sensing and identification system structure,including hardware design,system design and algorithm introduction,and then system implementation and evaluation,and finally put forward the application of the applica-tion-smoking behavior detection applications,the organization related experiments.The results of this thesis are as follows:1.The thesis designs a set of wearable devices based on smart necklaces and collects Non-Speech Body Sounds based on this device.The device is light and easy to wear,can be used in daily life.2.The thesis designs and implements the Non-Speech Body Sounds’ perception and recognition systems,including eating,drinking,breathing,breathing,sighing,cough-ing,and breathing.Through the wearable equipment to perceive the data,our sys-tem can process data hierarchically,classify data correctly.The results show com-pared to the previous work,our system has a certain higher accuracy.3.The thesis implements a demonstration application of the Non-Speech Body Sounds-smoking behavior detection to prove the value of our system,Our system collects the non-voice sound to detect deep breath of smoke,combined with the gestures and the voice of the lighter to detect smoking behavior.4.The system collected daily life Non-Speech Body Sounds and a number of smoking and real smokers daily life data to organize the experiment,the experimental results show that the system has a high accuracy,can be used in daily life.