Wearable Sensor Classification With Edge Offloading for Battery Life

The system segments the classification task by deploying multiple classifiers across multiple devices rather than relying on a single classifier on one device. Each device executes a subset of classifiers locally, and the results are aggregated to form the final classification decision. This segmentation allows each device to maintain fast local processing while collectively achieving higher accuracy through ensemble methods.

The patent merges the computational capabilities of multiple devices by aggregating predictions from multiple classifiers executed on different devices. The system combines results through a voting mechanism or other aggregation functions, effectively creating a distributed ensemble classifier that achieves both speed (through parallel processing) and accuracy (through multiple models).

The system implements local quality by allowing each device to execute only the classifiers relevant to its function or data type, rather than requiring all devices to run all classifiers. This reduces the complexity burden on individual devices while maintaining overall system reliability through the distributed architecture where failure of one device does not compromise the entire system.

The distributed classifier architecture provides beforehand cushioning by ensuring that the system can continue operating even if individual devices fail. Multiple classifiers are deployed across multiple devices, so the failure of one device only affects a portion of the classification task, not the entire system. This redundancy cushions against device failures and maintains system reliability.