IBIS: A Powerful Hybrid Architecture for Human Activity Recognition

TL;DR

This work tackles generalization in Wi-Fi sensing-based HAR by proposing IBIS, a hybrid network that combines spatial feature learning (Inception) with temporal modeling (BiLSTM) and an SVM-based post-processing step with an RBF kernel to refine decision boundaries. It demonstrates near 99% accuracy on Doppler-derived CSI data and shows clear improvements over SHARP and CNN-ABLSTM baselines, particularly in challenging generalization scenarios. The work highlights the benefit of integrating ensemble learning with Doppler features to achieve robust HAR in diverse environments, enabling privacy-preserving, low-cost sensing applications. It also analyzes computational trade-offs and suggests future work to reduce complexity while maintaining high accuracy.

Abstract

The increasing interest in Wi-Fi sensing stems from its potential to capture environmental data in a low-cost, non-intrusive way, making it ideal for applications like healthcare, space occupancy analysis, and gesture-based IoT control. However, a major limitation in this field is the common problem of overfitting, where models perform well on training data but fail to generalize to new data. To overcome this, we introduce a novel hybrid architecture that integrates Inception-BiLSTM with a Support Vector Machine (SVM), which we refer to as IBIS. Our IBIS approach is uniquely engineered to improve model generalization and create more robust classification boundaries. By applying this method to Doppler-derived data, we achieve a movement recognition accuracy of nearly 99%. Comprehensive performance metrics and confusion matrices confirm the significant effectiveness of our proposed solution.

Figures (8)

• Figure 1: The sequential process diagram illustrates the IBIS steps: collection of raw CSI data, passing through sanitization, neural network training, and Ensemble Learning post-Processing.
• Figure 2: Architecture of the IBIS hybrid neural network for Human Activity Recognition using Doppler traces, showing all convolutional and attention layers.
• Figure 3: Representation of the SVM algorithm applied to five movements with non-linear Data, applying the PCA method for dimensionality reduction.
• Figure 4: Comparative analysis of the confusion matrix for five movements: original Inception versus the hybrid IBIS with SVM post-processing.
• Figure 5: ROC curve analysis for five movements: comparing the original Inception with the hybrid Neural Network IBIS with SVM Post-Processing.