BTS: Bifold Teacher-Student in Semi-Supervised Learning for Indoor Two-Room Presence Detection Under Time-Varying CSI

TL;DR

A bifold teacher-student (BTS) learning approach for indoor human presence detection in a scenario with two adjoining rooms that outperforms existing SSL-based models in terms of the highest detection accuracy of around 98% while achieving the asymptotic performance of SL-based methods.

Abstract

In recent years, indoor human presence detection based on supervised learning (SL) and channel state information (CSI) has attracted much attention. However, existing studies that rely on spatial information of CSI are susceptible to environmental changes which degrade prediction accuracy. Moreover, SL-based methods require time-consuming data labeling for retraining models. Therefore, it is imperative to design a continuously monitored model using a semi-supervised learning (SSL) based scheme. In this paper, we conceive a bifold teacher-student (BTS) learning approach for indoor human presence detection in an adjoining two-room scenario. The proposed SSL-based primal-dual teacher-student network intelligently learns spatial and temporal features from labeled and unlabeled CSI datasets. Additionally, the enhanced penalized loss function leverages entropy and distance measures to distinguish drifted data, i.e., features of new datasets affected by time-varying effects and altered from the original distribution. Experimental results demonstrate that the proposed BTS system accomplishes an averaged accuracy of around 98% after retraining the model with unlabeled data. BTS can sustain an accuracy of 93% under the changed layout and environments. Furthermore, BTS outperforms existing SSL-based models in terms of the highest detection accuracy of around 98% while achieving the asymptotic performance of SL-based methods.

Figures (20)

• Figure 1: System architecture for human presence detection in adjacent rooms.
• Figure 2: Amplitude response of CSI (1) before and (2) after normalization. Note that we consider two TX and two RX antennas generating $K=4$ transmission pairs, labeled as $k\in\{1,2,3,4\}$ for respective subplots at top-left, top-right, bottom-left and bottom-right. The x-axis indicates the index of $S=52$ subcarrier, whilst y-axis is value of CSI amplitude.
• Figure 3: CSI observation of four cases of human presence. (a) Case 1: Empty in both rooms (b) Case 2: A person in room A and empty in room B (c) Case 3: Empty in room A and a person in room B (d) Case 4: people in both rooms. Note that we consider two TX and two RX antennas generating $K=4$ transmission pairs, labeled as $k\in\{1,2,3,4\}$ for respective subplots at top-left, top-right, bottom-left and bottom-right. The x-axis indicates the index of $S=52$ subcarrier, whilst y-axis is value of CSI amplitude.
• Figure 4: Amplitude response of normalized CSI in six different timestamps: (a) Round 1, (b) Round 2, (c) Round 3, (d) Round 4 with heavy rain, (e) Round 5, and (f) Round 6 with different antenna orientation. We consider Case 1 of empty room as an example. Note that we consider two TX and two RX antennas generating $K=4$ transmission pairs, labeled as $k\in\{1,2,3,4\}$ for respective subplots at top-left, top-right, bottom-left and bottom-right. The x-axis indicates the index of $S=52$ subcarrier, whilst y-axis is value of CSI amplitude.
• Figure 5: Schematic diagram of the proposed BTS system.