Non-Invasive Blood Glucose Meter
DOI:
https://doi.org/10.46243/jst.2021.v6.i04.pp67-74Keywords:
ML, ML.Ne, WiFiAbstract
The glucose concentration in the range from 111-330 mg/dl represents a diabetic patient. it causes physical impairment of vital organs. presently, non-invasive blood glucose monitoring device that uses the blood dielectric and impedance properties is becoming popular for diabetes detection. a realistic reusable low-cost non-invasive glucose meter is developed using wi-fi module. it measures glucose level based on the permittivity changes in human blood with sub-carriers. each sub-carrier is spaced with 312.5 khz bandwidth, generating 64 amplitudes and phases is used for channel estimation. a hampel filter is used to suppress abrupt amplitude variations occurring due to the environment effects. the system is trained using ml.net fast forest quantile regression method. the accuracy obtained is around 90%.