The invention discloses a low-power-consumption card detection intelligent awakening method based on
machine learning, which belongs to the field of low-power-consumption embedded application and comprises the following five steps: step 1, establishing an awakening and non-awakening dichotomy discrimination model; 2, collecting a
data set containing information such as an RSSI
time sequence; step 3, concluding and extracting features from the directions of
time domain,
frequency domain, difference, starting and ending features and the like; 4, dividing a
training set and a
test set for model training and performance evaluation respectively, and training a
decision tree model on a PC by utilizing
MATLAB to verify the accuracy; step 5, saving node information of the
decision tree in a FLASH of the MCU, acquiring the node information when the MCU is powered on, and controlling a lightweight advanced
rule engine of the card reading
chip through a configuration register so as to realize LPCD awakening; and step 6, continuous learning. The
system block diagram is shown in the figure 1, card approaching and various interference events are accurately distinguished, the false wake-up or non-wake-up rate of the
system is reduced, and the reliability is improved while the
power consumption of equipment is ensured.