Meta-Learning-Based People Counting and Localization Models Employing CSI from Commodity WiFi NICs

TL;DR

This work tackles reliable people counting and localization in indoor environments using CSI from commodity WiFi NICs, addressing offset and interference challenges with a low-latency CSI preprocessing pipeline. It then leverages model-agnostic meta-learning (MAML) to initialize and adapt CNN-based counting and localization models to new environments using only a small adaptation set. Empirical results across small/large rooms and open spaces show that Meta-CSI substantially outperforms non-preprocessed and transfer-learning baselines in both classification (sector/count) and regression (coordinates) tasks, often with only dozens of adaptation samples. The approach enables scalable, cross-environment sensing on standard hardware, though future work could explore unsupervised learning and nonlinear offset modeling to further improve robustness.

Abstract

In this paper, we consider people counting and localization systems exploiting channel state information (CSI) measured from commodity WiFi network interface cards (NICs). While CSI has useful information of amplitude and phase to describe signal propagation in a measurement environment of interest, CSI measurement suffers from offsets due to various uncertainties. Moreover, an uncontrollable external environment where other WiFi devices communicate each other induces interfering signals, resulting in erroneous CSI captured at a receiver. In this paper, preprocessing of CSI is first proposed for offset removal, and it guarantees low-latency operation without any filtering process. Afterwards, we design people counting and localization models based on pre-training. To be adaptive to different measurement environments, meta-learning-based people counting and localization models are also proposed. Numerical results show that the proposed meta-learning-based people counting and localization models can achieve high sensing accuracy, compared to other learning schemes that follow simple training and test procedures.

Figures (15)

• Figure 1: Structure of a CSI measurement system using commodity WiFi NICs for people counting and localization.
• Figure 2: A remote server conducts preprocessing of CSI and people counting and localization. CSI preprocessing includes offset removal of amplitudes and phases for each packet, and a learning model derives a sensing result based on the training and test sample sets of preprocessed CSI.
• Figure 3: A basic CNN module for the proposed people counting and localization models.
• Figure 4: A pre-training-based model using preprocessed CSI as model inputs. The pre-training-based model employs a simple CNN module, whose parameters are optimized through pre-training and adaptation phases.
• Figure 5: Procedure of CSI preprocessing. From erroneous CSI, a distinct feature of a given environment is extracted by reducing randomly generated offsets.