Indoor Passive Human Behavior Recognition Using WiFi Signal Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current human behavior recognition systems using radio frequency signals face low recognition accuracy due to differences in data distribution, despite advancements in deep learning and the use of CNN models with transfer learning and MMD measurement, which are hindered by issues like over-fitting and poor data distribution alignment.
Innovation Solution
An indoor passive human behavior recognition method and device employing transfer learning, where an indoor activity space is divided into regions, CIR data packets are collected and preprocessed, features are extracted, and a trained CNN model is used to recognize human behavior, incorporating techniques like filtering, wavelet transform, PCA, and MKMMD to improve accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If CNN model with transfer learning and MMD measurement is used for human behavior recognition, then the system can process wireless signal data automatically, but the recognition accuracy remains low due to data distribution differences and over-fitting
Solution Approach 1:
The patent segments the indoor activity space into multiple regions and divides the training data into domain-specific subsets (e.g., different rooms, different activity zones). This segmentation allows the model to learn region-specific patterns while reducing the negative impact of global data distribution differences, thereby improving recognition accuracy without sacrificing automation.
Solution Approach 2:
The patent modifies the MMD measurement parameter by introducing a region-aware weighting mechanism that dynamically adjusts the contribution of different regional data distributions. This parameter change enables the model to focus on critical regions while tolerating variations in less important areas, resolving the contradiction between automation and accuracy.
2Adaptability or versatility
If deep learning features are extracted from indoor wireless signals, then human activity recognition can be performed, but the data distribution differences cause over-fitting and low accuracy
Solution Approach 1:
The patent performs preliminary region-based data preprocessing and statistical analysis before training the deep learning model. By pre-identifying regional characteristics and adjusting data distributions in advance, the model avoids over-fitting to specific environments while maintaining versatility across different indoor settings.
Solution Approach 2:
The patent introduces region-based statistical features as an intermediary layer between the raw wireless signals and the CNN model. This intermediary transforms the data into a standardized format that reduces distribution differences while preserving activity-specific patterns, thereby improving both adaptability and reliability.
3Measurement precision
If traditional sensors and cameras are used for behavior recognition, then direct observation is possible, but wearing restrictions and system complexity increase
Solution Approach 1:
The patent replaces mechanical sensors and cameras with wireless signal-based detection. By using existing WiFi infrastructure to capture reflections and multipath effects caused by human activities, the system achieves behavior detection capability without adding physical sensors or cameras, thereby reducing device complexity while maintaining measurement precision.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The method significantly enhances recognition accuracy by effectively utilizing MIMO technology, reducing dimensionality, and improving communication quality, while being simple, cost-effective, and privacy-friendly, with high temporal and frequency resolution provided by DWT and PCA.
Implementation Method 1
radio frequency signals propagate in wireless media through multiple paths to reflect different objects and reach the receiver
Implementation Method 2
radio frequency signals propagate in wireless media through multiple paths to reflect different objects
Implementation Method 3
performing wavelet transform on the CIR data packet of the smoothed H (M, N, Z) matrix to obtain a denoised H (M, N, Z) matrix
Implementation Method 4
performing data dimension reduction processing on the CIR data packet of the denoised H (M, N, Z) matrix using PCA (Principal Component_analysis), so as to obtain a dimension-reduced H (M, N, Z) matrix
Data Source
AI summary
Disclosed are an indoor passive human behavior recognition method and device. The method includes the following steps: dividing an indoor activity space into multiple regions, collecting a channel impulse response data packet of a reflection signal of each activity in each region to obtain an H (M, N, Z) matrix; preprocessing the H (M, N, Z) matrix to obtain a preprocessed H (M, N, Z) matrix; extracting features of the preprocessed H (M, N, Z) matrix to obtain a training sample of a convolutional neural network model; performing transfer learning on the convolutional neural network model using the training sample to obtain a trained convolutional neural network model; obtaining an indoor channel impulse response amplitude value, inputting the channel impulse response amplitude value into the trained convolutional neural network model, and outputting a human behavior.