Method and system for updating machine learning based classifiers for reconfigurable sensors

An electronic device, comprising:

• a sensor unit, comprising:
• an inertial sensor configured to generate sensor signals;
• an amplifier coupled to the inertial sensor and configured to amplify the sensor signals;
• an analog to digital converter configured to generating digital sensor data by converting the amplified sensor signals to digital sensor data;
• an arithmetic logic unit configured to receive the digital sensor data and to generate, based on the sensor data, feature data including a first set of features;
• an interrupt generator configured to generate an interrupt based on the feature data received from the arithmetic logic unit;
• a memory configured to store the sensor data with a current back of sensor data;
• a centroid buffer configured to store a previous centroid value associated with a previous batch of sensor data generated by the inertial sensor;
• a data checking module configured to receive a divergence score relative to a centroid value associated with the feature data and the previous centroid value;
• a transceiver configured to transmit the batch of sensor data from the electronic device responsive to the divergence score;
• a configuration register coupled to the arithmetic unit and configured to receive feature configuration data responsive to the transceiver transmitting the batch of sensor data, wherein:
• the inertial sensor is configured to generate new sensor signal after the configuration register receives the feature configuration data,
• the analog to digital converter is configured to generate new sensor data based on the new sensor signals;
• and
• the arithmetic logic unit is configured to generate new feature data based on the new sensor data and including a second set of features different than the first set based on the feature configuration data.

A method, comprising:

• receiving, with a staging buffer, inertial sensor data from an inertial sensor of a sensor unit of an electronic device;
• receiving, with the staging buffer, configuration data of the sensor unit including operating parameters of the inertial sensor;
• receiving, with the stage buffer from an arithmetic logic unit, feature data based on the inertial sensor data and the configuration data and including a plurality of features;
• storing, in the staging buffer, the sensor data with a current batch of sensor data;
• receiving a centroid value based on the feature data;
• storing, in a centroid buffer, a previous centroid value associated with a previous batch of sensor data generated by the inertial sensor;
• receiving a divergence score relative to the centroid value and the previous centroid value associated with previous feature data of the previous batch of sensor data;
• transferring the sensor data from the staging buffer to a global buffer responsive to the divergence score;
• generating, responsive to the divergence score, new configuration data including new operating parameters of the inertial sensor based on the sensor data;
• transmitting, to the electronic device, the new configuration data responsive to the divergence score;
• receiving with the staging buffer, new inertial sensor data from the inertial sensor of the sensor unit of the electronic device based on the new configuration data after transmitting the new configuration data;
• and
• receiving, with the staging buffer, new feature data based on the new inertial sensor data and the new configuration data.