Sense+ — Adaptive signal preprocessing
A plug-and-play preprocessing module for human-centered wireless sensing.
The problem
Wireless sensing relies on preprocessing to reduce noise and reveal task-relevant information. However, pipelines tailored to a single signal type or task can be difficult to reuse.
The approach
Sense+ is a plug-and-play module that integrates task-adaptive low-pass filtering into neural network pipelines. Motivated by motion-related frequency shifts, it learns filter weights for the downstream sensing task.
Results
When integrated into a neural network, the module improves gesture recognition accuracy by 21.89%.
Publication
Sense+: A Plug-and-Play Signal Preprocessing Approach for Enhancing Human-Centered Wireless Sensing. IEEE Internet of Things Journal, vol. 12, no. 19, pp. 40531–40544, Oct. 2025. DOI.