A Research On Liquid Remaining Volume Detection Technology Based On Smart Phone Sensors

Nowadays,with the increasing pressure of people's life and work,people's health problems are becoming more and more prominent.According to statistics,about 70% of the Chinese people are in a sub-healthy state,so it becomes more and more important to pay attention to their health.At present,the concept of health management is gradually coming into people's view,and the domestic market of health management is also grad-ually hot.A more important part of health management is the balance of nutrient intake,and the daily nutrient required by the human body is ingested through liquids,such as wa-ter,vitamins,protein,etc.,so liquid detection technology appears to be a important way of nutrition monitoring.Liquid detection technology can help people detect the intake of milk,juice,water and other liquids by detecting the remaining amount of liquid,and provide relevant support for health management and accurate recommendations of busi-nesses.Based on the smartphone sensor,this thesis proposes an effective and feasible system for the liquid detection problem of how to detect the remaining volume of barreled milk,which we call DeMilk.The core idea of DeMilk is to detect the remaining volume of milk by identifying the difference in the sound of the different remaining volume of milk.DeMilk is achieved through the following 4 steps:(1)Data collection.Collection of shaking sound data for different residual amounts of milk in barrels and normalization of the shaking milk pose by means of linear accelerometers and gyroscopes,followed by data segmentation and pre-processing for data noise reduction.?(2)feature extraction.Extract the corresponding features from each sample data?(3)Feature filter.Filter the extracted features through different methods?(4)Model training.The model is trained based on the collected data set and the optimal feature set obtained by filter.The main innovations and research results of this thesis are as follows:(1)This thesis is the first to detect the amount of liquid remaining by fusing a smart-phone linear accelerometer,gyroscope and microphone sensor.(2)A total of 26 features were extracted for the detection of liquid residuals in this thesis.The features are based on six aspects: short-time Fourier transform,short-time average over-zero rate,spectral centroid,wavelet transform,Mel-scale frequency cepstral coefficients and onset detection.(3)The support vector machine model generated by the research in this thesis has the best effect,with an accuracy of 0.9240,an F1 score of 0.9208,and an AUC value of0.9857.(4)This thesis implements the entire DeMilk system as an Android application.