Sensor-Aware Classifiers for Energy-Efficient Time Series Applications on IoT Devices
Dina Hussein, Lubah Nelson, Ganapati Bhat
TL;DR
This work tackles the high sensing energy cost of time-series IoT classification by introducing Sensor-Aware Early Exit (SEE), which processes only partial sensor windows with early exits and uses late-input blocks to incorporate additional data as needed. SEE is instantiated for both 1-D CNNs and ensembles of random forests, trained with a multi-exit loss and inference guided by entropy-based confidence thresholds to decide when to stop sensing. Across six health/activity datasets on edge hardware, SEE achieves 50–60% average energy savings with accuracy comparable to full-window baselines, demonstrating strong practical potential for energy-constrained IoT applications. The approach is validated through a design-space exploration that tunes exit placements, data fractions, and thresholds, showing robust performance and scalable applicability to real-world time-series sensing tasks.
Abstract
Time-series data processing is an important component of many real-world applications, such as health monitoring, environmental monitoring, and digital agriculture. These applications collect distinct windows of sensor data (e.g., few seconds) and process them to assess the environment. Machine learning (ML) models are being employed in time-series applications due to their generalization abilities for classification. State-of-the-art time-series applications wait for entire sensor data window to become available before processing the data using ML algorithms, resulting in high sensor energy consumption. However, not all situations require processing full sensor window to make accurate inference. For instance, in activity recognition, sitting and standing activities can be inferred with partial windows. Using this insight, we propose to employ early exit classifiers with partial sensor windows to minimize energy consumption while maintaining accuracy. Specifically, we first utilize multiple early exits with successively increasing amount of data as they become available in a window. If early exits provide inference with high confidence, we return the label and enter low power mode for sensors. The proposed approach has potential to enable significant energy savings in time series applications. We utilize neural networks and random forest classifiers to evaluate our approach. Our evaluations with six datasets show that the proposed approach enables up to 50-60% energy savings on average without any impact on accuracy. The energy savings can enable time-series applications in remote locations with limited energy availability.