The primary challenge was to develop a system that could harness commodity WiFi signals to provide real-time spatial intelligence, including presence detection, vital sign monitoring, and activity recognition, all without relying on cameras or wearables. This required overcoming significant technical hurdles, such as accurately capturing and interpreting WiFi signal disturbances caused by human presence and movement.

• Developing a robust and efficient algorithm for processing Channel State Information (CSI) from low-cost ESP32 sensors.
• Creating a scalable and adaptable system that could learn each environment locally using spiking neural networks and multi-frequency mesh scanning.
• Ensuring the system's accuracy and reliability in various settings, including through-wall sensing and multi-person detection.

The solution involved designing and implementing a comprehensive system comprising several key components, including:

• ESP32 Sensors: Utilizing low-cost ESP32 boards to capture CSI from WiFi signals, which are then processed to extract meaningful information about the environment.
• Spiking Neural Networks: Implementing spiking neural networks that can learn and adapt to each environment in under 30 seconds, enabling real-time spatial intelligence.
• Multi-Frequency Mesh Scanning: Developing a multi-frequency mesh scanning approach that leverages neighboring WiFi routers as free radar illuminators to enhance sensing capabilities.
• Cognitum Seed: Integrating the Cognitum Seed for persistent memory, cryptographic attestation, and AI integration, ensuring the system's security and reliability.
• Pre-Trained Models: Utilizing pre-trained models, such as the wifi-densepose-pretrained model, to enable accurate presence detection, vital sign monitoring, and activity recognition. The system's architecture is designed to be modular and scalable, with a focus on edge computing and real-time processing. This allows for instant response times and eliminates the need for cloud connectivity or internet access.

import numpy as np from sklearn.neighbors import KNeighborsClassifier

Load pre-trained model

model = KNeighborsClassifier(n_neighbors=5)

Load CSI data from ESP32 sensor

csi_data = np.load('csi_data.npy')

Process CSI data and extract features

features = process_csi_data(csi_data)

Use pre-trained model for presence detection and vital sign monitoring

predictions = model.predict(features)