Joint Activity Recognition and Indoor Localization with WiFi Fingerprints (1904.04964v2)

Abstract: Recent years have witnessed the rapid development in the research topic of WiFi sensing that automatically senses human with commercial WiFi devices. This work falls into two major categories, i.e., the activity recognition and the indoor localization. The former work utilizes WiFi devices to recognize human daily activities such as smoking, walking, and dancing. The latter one, indoor localization, can be used for indoor navigation, location-based services, and through-wall surveillance. The key rationale behind this type of work is that people behaviors can influence the WiFi signal propagation and introduce specific patterns into WiFi signals, called WiFi fingerprints, which can be further explored to identify human activities and locations. In this paper, we propose a novel deep learning framework for joint activity recognition and indoor localization task using WiFi Channel State Information~(CSI) fingerprints. More precisely, we develop a system running standard IEEE 802.11n WiFi protocol, and collect more than 1400 CSI fingerprints on 6 activities at 16 indoor locations. Then we propose a dual-task convolutional neural network with 1-dimensional convolutional layers for the joint task of activity recognition and indoor localization. Experimental results and ablation study show that our approach achieves good performances in this joint WiFi sensing task. Data and code have been made publicly available at https://github.com/geekfeiw/apl.

Joint Activity Recognition and Indoor Localization with WiFi Fingerprints

The paper "Joint Activity Recognition and Indoor Localization with WiFi Fingerprints" provides an examination of utilizing WiFi Channel State Information (CSI) to simultaneously achieve activity recognition and indoor localization. This dual-purpose approach leverages the ubiquity of commercial WiFi devices to enhance human-computer interaction applications, particularly within smart home environments.

Overview of the Methodology

The researchers propose a deep learning framework featuring a dual-task convolutional neural network using one-dimensional convolutional layers to manage both tasks. The experimental setup exhibits IEEE 802.11n protocol employed on USRP (Universal Software Radio Peripheral) devices, collecting over 1400 CSI fingerprints that correspond to six distinct gestures across sixteen distinct locations.

The neural network architecture, termed ResNet1D, involves cascading multiple residual blocks in a format resembling the well-known ResNet, tailored for one-dimensional temporal data. The network's task is to predict both activity and location labels concurrently, examining how CSI fingerprints change with each activity and location.

Experimental Results

The results achieved were notable, with activity recognition accuracy reported at 88.13% and indoor localization accuracy at 95.68%. The network was also analyzed visually using t-SNE to understand how well the deep learning model differentiates between activities and locations through feature extraction. The data suggests that the model effectively processes temporal variations intrinsic to CSI data representing distinct activities and locations.

To further analyze performance, quantitative measures such as precision, recall, and F1 scores were assessed, revealing the model's strong performance across most classes with some variance in specific activities like 'hand circle'. Notably, deeper network configurations improved localization accuracy but exhibited decreasing returns on activity recognition.

Implications and Future Directions

This research highlights the practical utility of using existing WiFi infrastructure for activity and location recognition, reducing the need for additional hardware. It opens avenues for enhanced IoT capabilities within smart homes, enabling devices to understand context through both user actions and their locations. Such insights could significantly improve interaction personalization and the automation of context-aware services.

For future work, extending this dual-task approach to more complex environments or varying device configurations could provide a comprehensive view of its robustness. Potential expansions might also explore real-time deployments or algorithm optimizations to reduce computational demands, supporting broader applicability in resource-constrained settings.

In conclusion, the proposed method showcases a compelling fusion of WiFi sensing and deep learning, illustrating a promising step towards sophisticated contextual interaction within connected spaces. As the integration of AI and IoT technologies continues to grow, methodologies like those explored in this paper could play vital roles in shaping the trajectories of smart environments.