Mel-Frequency Cepstral Coefficients (MFCC) is a feature extraction method originally developed for speech processing that transforms a time-domain signal into a compact set of coefficients representing the spectral envelope of the signal on a perceptually motivated mel frequency scale. In WiFi-based human sensing, MFCCs are applied to CSI or RSSI time-series data to capture discriminative temporal and frequency characteristics of human activities, making them particularly useful for gesture recognition, activity classification, and gait analysis tasks where compact yet expressive representations are needed. Key variants in this context include the use of delta and delta-delta MFCCs, which incorporate first- and second-order temporal derivatives to encode dynamic changes in the signal, further enriching the feature set before input into machine learning or deep learning classifiers.
Source Papers
- A Survey on Wireless Device-free Human Sensing: Application Scenarios, Current Solutions, and Open Issues ↗ — A Survey on Wireless Device-free Human Sensing: Application
- Towards Environment Independent Device Free Human Activity Recognition ↗ — Towards Environment Independent Device Free Human Activity R
- WiFi-Based Human Sensing With Deep Learning: Recent Advances, Challenges, and Opportunities ↗ — WiFi-Based Human Sensing With Deep Learning: Recent Advances