Project H.A.W.K.

Human Activity Detection Through Walls Using Microwave Kinetics

A FreeRTOS-powered through-wall life detection radar with real-time vital sign extraction, distance estimation, and a tactical LAN dashboard.

Overview

Project H.A.W.K. is an RTOS-based embedded system engineered for non-line-of-sight (NLOS) human detection using microwave Doppler radar. The system penetrates solid obstacles — drywall, rubble, wood, and smoke — to detect physiological micro-movements: breathing (0.2–0.6 Hz) and heartbeat (1.0–2.5 Hz).

Built on an ESP32 dual-core processor running FreeRTOS, the architecture guarantees deterministic timing and real-time preemptive multitasking across four concurrent tasks. A 1024-point FFT pipeline with parabolic interpolation provides sub-bin frequency accuracy for vital sign extraction. A 15-second auto-calibration sequence dynamically learns the environmental noise floor on every boot using 75th-percentile statistical analysis for outlier resilience, and all telemetry is streamed over WebSocket to a tactical browser-based dashboard accessible from any device on the local network.

Key Features

Through-wall human detection via 10.525 GHz microwave Doppler radar
Dual-band vital sign extraction — breathing rate (BPM) and heart rate (BPM)
Distance estimation — estimates target range using empirical signal magnitude decay
Dynamic auto-calibration — 75th-percentile noise floor analysis, resilient to boot-time interference
Confidence-based detection — 5-step ramp-up with gentle decay (−1) for reliable slow-breather detection
Fast re-acquisition — post-alert confidence starts at 3/5, requiring only ~8s to re-confirm
Real-time tactical dashboard — WebSocket-based, military-themed, with live charts, BPM display, range estimation, and signal strength monitoring

Key Applications

Domain	Use Case
Tactical & Defense	Urban warfare situational awareness — detect human presence behind walls before room entry, hostage rescue, and tactical reconnaissance.
Search & Rescue	Locate survivors trapped under earthquake debris or collapsed structures by detecting faint vital signs through rubble.
Advanced Security	Covert perimeter monitoring and intrusion detection in zero-visibility environments (smoke, darkness, fog).
Medical Monitoring	Non-contact vital sign monitoring in isolation wards or hazardous environments.

System Architecture

+-----------------------------------------------------------------------------+
|                     PROJECT H.A.W.K. -- SYSTEM OVERVIEW                     |
+-----------------------------------------------------------------------------+
|                                                                             |
|  +--------------+    +--------------+    +----------------------------+     |
|  | HB100/CDM324 |    | Active BPF   |    |    ESP32 (Dual-Core)       |     |
|  | Doppler      |--->| 0.1-3.0 Hz   |--->|     FreeRTOS Kernel        |     |
|  | Radar        | IF | Op-Amp       | ADC|                            |     |
|  | 10.525 GHz   |    | Circuit      |    | +------------------------+ |     |
|  +--------------+    +--------------+    | | T1: Radar Sampling     | |     |
|                                          | |     250 Hz (P4)        | |     |
|                                          | +------------------------+ |     |
|                                          | | T2: 1024-pt FFT        | |     |
|  +--------------------------------------+| |     Pipeline (P3)      | |     |
|  |     TACTICAL LAN DASHBOARD           || +------------------------+ |     |
|  | +--------+ +--------+ +------------+ || | T3: Detection &        | |     |
|  | | Vital  | | Range  | | Event      | |<-|     Calibration (P2)   | |     |
|  | | Charts | | & BPM  | | Log        | || +------------------------+ |     |
|  | +--------+ +--------+ +------------+ || | T4: WebSocket &        | |     |
|  | WebSocket on ws://<ESP32_IP>:81      || |     Alarms (P1)        | |     |
|  +--------------------------------------+| +------------------------+ |     |
|                                          +----------------------------+     |
+-----------------------------------------------------------------------------+

Hardware Architecture

Component	Specification	Purpose
Microcontroller	ESP32 DevKit (32-bit, Dual-Core Xtensa LX6, 240 MHz)	FreeRTOS host, ADC sampling, WiFi/WebSocket server
Radar Sensor	HB100 (10.525 GHz) / CDM324 (24 GHz) Microwave Doppler	Emits CW microwave; IF output encodes target micro-motion
Signal Conditioning	Custom 2-stage Active Bandpass Filter (LM358 Op-Amp)	Passband: 0.1–3.0 Hz — isolates human vitals, rejects mechanical noise
ADC Interface	ESP32 12-bit SAR ADC (GPIO 34)	0–4095 counts at 250 Hz deterministic sampling
Alert Output	Active Buzzer (GPIO 25) + LED Indicator (GPIO 2)	1.5-second local hardware alarm on confirmed detection
Dashboard	Any browser on the LAN	Real-time tactical display via WebSocket (`ws://<IP>:81`)

Pin Map

Signal	GPIO	Direction
Radar ADC (Bandpass Filter Output)	`34`	Input (Analog)
Active Buzzer	`25`	Output (Digital)
Status LED	`2`	Output (Digital)

FreeRTOS Task Architecture

The system employs a preemptive priority-based RTOS scheduler with four concurrent tasks communicating through FreeRTOS Queues and Semaphores. The Arduino loop() is immediately deleted to free CPU cycles for the RTOS kernel.

Task Overview

#	Task	Priority	Stack	Function	IPC Mechanism
T1	Radar Acquisition	4 (Highest)	2 KB	250 Hz deterministic ADC sampling via `vTaskDelayUntil()`	Produces → `rawDataQueue` (2048 × `uint16_t`)
T2	Signal Processing	3	8 KB	1024-point FFT (DC removal → Hamming window → FFT → Magnitude → Sub-band peak detection)	Consumes `rawDataQueue` → Produces → `processedDataQueue` (8 × `VitalSignData`)
T3	Detection & Calibration	2	4 KB	Auto-calibration (15 s, 75th-percentile) + confidence algorithm + distance estimation + BPM conversion	Consumes `processedDataQueue` → Produces → `dashboardQueue` + `detectionSemaphore`
T4	Communications & UI	1 (Lowest)	8 KB	WebSocket broadcast at ~20 Hz loop, buzzer/LED alarm with 1.5s auto-reset	Consumes `dashboardQueue` / `detectionSemaphore`

Inter-Task Data Flow

  rawDataQueue            processedDataQueue          dashboardQueue
   (2048 slots)              (8 slots)                  (4 slots)
       |                         |                          |
T1 --->|--> T2 ----------------->|--> T3 ------------------>|--> T4 --> WebSocket
(250 Hz)    (1024-pt FFT)             (Calibration +             (Broadcast +
                                       Confidence +               Buzzer/LED)
                                       Distance Est.)
                                          |
                                          v
                                 detectionSemaphore
                                      (Binary)

Signal Processing Pipeline

Each FFT cycle processes 1024 samples collected over ~4.1 seconds at 250 Hz, yielding a frequency resolution of 0.2441 Hz/bin.

Pipeline Stages

DC Removal — Strips the ADC bias offset to center the waveform at zero.
Hamming Windowing — Reduces spectral leakage at bin boundaries.
1024-Point FFT — Full complex FFT using float (not double) to leverage the ESP32's hardware single-precision FPU (~30 ms vs ~300 ms on double).
Magnitude Extraction — complexToMagnitude() converts complex bins to real magnitudes.
Dual Sub-Band Peak Detection — Independent peak search with parabolic interpolation for sub-bin frequency accuracy:

Vital Sign	Frequency Band	FFT Bin Range	Resolution
Breathing	0.20 – 0.60 Hz	Bins 1–2	±0.12 Hz
Heartbeat	1.00 – 2.50 Hz	Bins 4–10	±0.12 Hz

BPM Conversion — Detected frequencies are converted to beats/breaths per minute (freq × 60).

Key Design Decisions

float over double: The ESP32's FPU is single-precision only. Using double forces software emulation at ~10x the compute cost — unacceptable for a real-time system.
static arrays: The 8 KB FFT buffers (vReal[1024], vImag[1024]) are declared static to place them in BSS rather than on the task stack, preventing stack overflow.
DC-safe bin clamping: Bin 0 (DC component) is always excluded from peak search, and parabolic interpolation results are clamped to the safe band range to prevent negative-frequency artifacts.

Dynamic Auto-Calibration

Instead of a hardcoded noise threshold, the system performs a 15-second environmental calibration on every boot using 75th-percentile statistical analysis for outlier resilience.

Calibration Sequence

BOOT --> STATE_CALIBRATING (15s) --> STATE_ACTIVE
              |                            |
              | Record peak FFT magnitude  | Use dynamic threshold:
              | for each FFT cycle into    |   activeThreshold =
              | a circular buffer          |     P75(noise) x 1.5
              |                            |
              | Compute 75th percentile    | Confidence algorithm
              | (resilient to outliers)    | fully operational
              |                            |
              v                            v
         Dashboard shows             Dashboard shows
         amber CALIBRATION           full tactical view
         overlay with                with live charts,
         progress bar                BPM, and distance

Parameter	Value	Purpose
`CALIBRATION_DURATION_MS`	15000 ms	Window for noise floor measurement
`SAFETY_MARGIN`	1.5x	Multiplier applied to 75th-percentile noise magnitude
`FALLBACK_THRESHOLD`	50.0	Used if calibration sees zero signal (dead ADC)
`REQUIRED_CONFIDENCE`	5	Consecutive positive readings needed for alert
`CONFIDENCE_DECAY`	1	Per-cycle decay on negative readings (gentle — supports slow breathers)
`POST_ALERT_CONFIDENCE`	3	Confidence level after alert (fast re-acquisition in ~8s)

Non-Blocking Design

The calibration phase uses xTaskGetTickCount() comparisons — no delay() calls — so the Detection Task continues consuming the processedDataQueue and feeding telemetry to the dashboard throughout the entire calibration window.

Why 75th Percentile Instead of Peak?

The original design tracked the absolute peak noise magnitude during calibration. This was vulnerable to poisoning — if a person stood near the sensor during boot, their vital signs would inflate the peak, making the threshold unreachably high. The 75th percentile approach uses std::sort on a buffer of calibration samples and selects the value at the 75% mark. This rejects the top 25% of readings (likely outliers) while still capturing the true noise floor.

Distance Estimation

The HB100 is a CW (Continuous Wave) radar and cannot perform time-of-flight ranging like FMCW radar. However, empirical measurements show that FFT signal magnitude decays predictably with distance, following an inverse power law:

magnitude ≈ A / (distance ^ n)

By inverting this relationship:

distance ≈ (A / net_magnitude) ^ (1/n)

Where:

A = 0.55 (amplitude coefficient, fitted from empirical data through drywall)
n = 1.8 (decay exponent)
net_magnitude = measured_magnitude - noise_floor

Empirical Validation

Distance (m)	Measured Magnitude	Detection Status
0.5	0.97	✅ Strong
1.0	0.68	✅ Strong
1.5	0.33	✅ Moderate
2.0	0.27	✅ Weak
2.5	0.12	⚠ At noise floor
3.0	0.05	❌ Below noise floor

Limitations

Distance estimates are signal-strength-based approximations, not precise measurements
Accuracy varies with barrier material (drywall vs. brick vs. plywood)
The A and n constants should be recalibrated for your specific environment
Reports N/A when signal is too weak for meaningful estimation

Tactical LAN Dashboard

A military-themed, browser-based monitoring console that connects to the ESP32 via WebSocket on port 81. No server, no cloud — just open index.html from any device on the same LAN.

Dashboard Features

Feature	Description
Connection Overlay	Full-screen animated radar sweep with live reconnect countdown
IP Configuration	Setup screen to enter the ESP32's IP address before connecting
Calibration Overlay	Amber warning overlay with progress bar and countdown timer during the 15s noise floor analysis
Threat Assessment	Real-time status panel — transitions from green `MONITORING` to red-alert `HUMAN DETECTED` with flashing vignette
Confidence Gauge	5-bar visual gauge (green to amber to red) with labels: NO SIGNAL, WEAK, TRACKING, ACQUIRING, LOCKED, CONFIRMED
Vital Sign Charts	Dual Chart.js line graphs — Breathing (0.2-0.6 Hz, cyan) and Heartbeat (1.0-2.5 Hz, magenta) — scrolling 30-point window
Vital Signs (BPM)	Real-time breathing rate and heart rate displayed in beats/breaths per minute
Estimated Range	Approximate distance to the detected human target (meters)
Noise Floor Monitor	Current noise floor magnitude for environmental diagnostics
Detection Duration	Timer showing how long a human has been continuously detected
Signal Strength	Raw FFT magnitude of the strongest vital-sign peak
System Metrics	Total alerts, current breathing Hz, current heartbeat Hz, last alert timestamp
Event Log	Scrolling console with timestamped telemetry, system events, and alerts (max 150 entries)
Audio Alert	Web Audio API tactical 3-tone square-wave alarm on human detection
Auto-Reconnect	Automatic WebSocket recovery with visual countdown on connection loss

Tech Stack

Layer	Technology
Structure	Semantic HTML5
Styling	Vanilla CSS — glassmorphism, CSS custom properties, CRT scan-line overlay
Typography	Orbitron (display) + JetBrains Mono (data) via Google Fonts
Charts	Chart.js 4.4.7 (CDN)
Data Transport	Native WebSocket API, `ws://<ESP32_IP>:81`

Communication Protocol

All telemetry is transmitted exclusively over WebSocket (Serial output is disabled in production firmware to eliminate UART jitter in the time-critical 250 Hz sampling loop).

$HAWK Protocol

Message Type	Format	Direction
Handshake	`$HAWK,STATUS,CONNECTED`	ESP32 to Dashboard
Telemetry	`$HAWK,DATA,<breathFreq>,<heartFreq>,<breathMag>,<heartMag>,<confidence>,<maxConf>,<timestamp>,<state>,<breathBPM>,<heartBPM>,<estRange>,<noiseFloor>,<detectDur>`	ESP32 to Dashboard
Alert	`$HAWK,ALERT,HUMAN_DETECTED,TIME:<ms_timestamp>`	ESP32 to Dashboard

Telemetry Fields

Field	Type	Description
`breathFreq`	`float`	Dominant frequency in breathing band (Hz)
`heartFreq`	`float`	Dominant frequency in heartbeat band (Hz)
`breathMag`	`float`	FFT magnitude of breathing peak (signal strength)
`heartMag`	`float`	FFT magnitude of heartbeat peak (signal strength)
`confidence`	`int`	Current detection confidence (0 to `maxConf`)
`maxConf`	`int`	Required confidence threshold for alert trigger
`timestamp`	`ulong`	ESP32 `millis()` uptime
`state`	`string`	`CALIBRATING` or `ACTIVE`
`breathBPM`	`float`	Breathing rate in breaths per minute
`heartBPM`	`float`	Heart rate in beats per minute
`estRange`	`float`	Estimated distance to target in meters (-1 = unknown)
`noiseFloor`	`float`	Current noise floor magnitude
`detectDur`	`ulong`	Duration of continuous detection (milliseconds)

Screenshots & Visual Documentation

System Block Diagram

Hierarchical tree diagram showing the 5 major subsystems of H.A.W.K.: Target Environment (human subject behind non-metallic barrier), Microwave Doppler Radar (10.525 GHz CW transmission, phase-shifted reflections, raw IF signal), Analog Signal Conditioning Circuit (LM358 dual op-amp, active bandpass filter, amplification & 1.65V DC bias), Microcontroller Processing (ESP32 with FreeRTOS — ADC sampling, 1024-pt FFT, peak detection, confidence algorithm), and Output & Telemetry (local buzzer/LED alarms, WebSocket server, tactical dashboard).

Hardware Prototype

Physical lab setup showing the complete H.A.W.K. system in operation. The ESP32 development board is visible on the breadboard (left), connected to the HB100 radar module (green PCB, right) through the op-amp signal conditioning circuit with jumper wires. The H.A.W.K. tactical dashboard is running live on the laptop in the background, displaying real-time breathing and heartbeat frequency charts, confidence gauge, and the scrolling event log.

LTspice Amplifier Circuit Schematic

Complete 2-stage active bandpass filter designed in LTspice using dual op-amps (U1, U2). Stage 1: C1 (150 µF) + R3 (10 kΩ) high-pass input → U1 with R4 (1 MΩ) ‖ C2 (0.047 µF) feedback. Stage 2: C3 (150 µF) + R5 (10 kΩ) → U2 with R6 (1 MΩ) ‖ C4 (0.047 µF) feedback. DC bias at 1.65V via R1/R2 (10 kΩ/10 kΩ) voltage divider from V3 (3.3V). Input source: SINE(0 100µ 0.3) simulating the HB100 IF signal. Output labeled "ESP32" connects to GPIO 34.

Bandpass Filter Frequency Response

Bode magnitude plot from LTspice AC analysis (.ac dec 100 0.01 10) showing the dual-stage filter's frequency response. The shaded passband spans 0.1 Hz (HPF cutoff) to 3.0 Hz (LPF cutoff) with a peak gain of approximately 78 dB near 0.5 Hz. The response rolls off cleanly on both sides, confirming the filter correctly isolates the human vital-sign frequency range while rejecting DC drift and high-frequency mechanical noise.

MATLAB Baseband Signal Analysis

Dual-panel MATLAB plot. Top: Live baseband signal in the time domain showing 10 seconds of raw ADC voltage centered at ~2.12V (DC bias from the op-amp circuit) with visible micro-amplitude modulations caused by human breathing. Bottom: Extracted vital signs via FFT in the frequency domain (0–3 Hz) with the breathing peak annotated at 0.20 Hz (red circle), confirming successful vital sign extraction from the radar signal.

ADC Sampling Determinism

MATLAB plot of inter-sample intervals across ~500 samples (sample indices ~25,385–25,434). The flat line at exactly 4.0 ms confirms the FreeRTOS vTaskDelayUntil() mechanism achieves perfect deterministic timing at 250 Hz with zero jitter — critical for accurate FFT frequency resolution.

Dynamic Auto-Calibration & Target Acquisition

MATLAB plot illustrating the full calibration-to-detection lifecycle. The shaded region (0–15 s) shows the calibration phase where peak noise magnitude is tracked (~18 units). At t = 15 s (dashed line), calibration ends and the dynamic threshold (θ_active = P75(noise) × 1.5 ≈ 27) is locked. Post-calibration, detected peak magnitudes jump to 70–90 units, clearly exceeding the threshold, demonstrating successful human detection. The label "ACTIVE (HUMAN DETECTED)" marks the operational phase.

Dashboard — Live Vital Sign Telemetry

MATLAB-generated dual-panel plot showing real-time vital sign extraction over a ~2-minute window. Top panel: Breathing frequency (0.2–0.5 Hz band) showing a characteristic baseline around 0.24 Hz with periodic spikes to ~0.49 Hz indicating active respiration cycles. Bottom panel: Heartbeat frequency (1.0–2.5 Hz band) showing oscillating peaks between 0.97 Hz and 2.44 Hz, demonstrating successful dual-band vital sign separation.

Signal Magnitude vs. Distance (Through Drywall)

Empirical measurement of breathing signal FFT peak magnitude as a function of distance from the radar sensor (through 12.5 mm drywall). Signal decays from ~0.97 at 0.5 m to ~0.05 at 3.0 m. The mean noise floor (dashed line at ~0.15) intersects the signal curve at approximately 2.5 m (circled), defining the maximum practical detection range through drywall for breathing detection. This data is used to calibrate the power-law distance estimation model.

Vital Sign Degradation by Barrier Material

Bar chart comparing normalized FFT peak magnitudes across 4 barrier types: Line of Sight (0.98), Drywall / 12.5 mm (0.74), Plywood / 18 mm (0.49), and Brick / 15 cm (0.17). Error bars show measurement variance across multiple trials. Demonstrates the radar's through-wall capability with quantified signal attenuation per material type.

RTOS Scheduling Timeline

WaveDrom-generated timing diagram illustrating FreeRTOS task scheduling and IPC queue interactions. Shows T1 (Acquisition, P4) running ADC samples at highest priority, T2 (Processing, P3) running FFT batches, T3 (Detection, P2) evaluating confidence, and T4 (UI, P1) broadcasting WebSocket packets. Below, the IPC queues (rawQ, procQ, dashQ) show data flow pulses between tasks. Priority order: P4 > P3 > P2 > P1.

Repository Structure

H.A.W.K/
|
+-- ESP32/                             # Embedded firmware (Arduino/FreeRTOS)
|   +-- ESP32.ino                      # Entry point -- Queue/Semaphore creation, WiFi init, task launch
|   +-- globals.h                      # Pin definitions, SystemState enum, VitalSignData struct, RTOS handles
|   +-- radar_sensor.cpp / .h          # T1: 250 Hz deterministic ADC sampling task
|   +-- signal_processing.cpp / .h     # T2: 1024-pt FFT pipeline with dual sub-band peak detection
|   +-- detection.cpp / .h             # T3: Auto-calibration (75th percentile) + confidence + distance estimation + BPM
|   +-- comms_ui.cpp / .h              # T4: WiFi/WebSocket server, $HAWK protocol (v2), buzzer/LED alarms
|
+-- Dashboard/                         # Browser-based tactical monitoring console
|   +-- index.html                     # Semantic HTML5 structure with all overlay layers + extended metrics
|   +-- styles.css                     # Full design system -- dark theme, glassmorphism, animations
|   +-- app.js                         # WebSocket client, Chart.js graphs, v2 telemetry parser, BPM/range display
|
+-- LTspice/                           # Analog circuit simulation files
|   +-- HB100-ESP32.asc                # LTspice schematic -- 2-stage active bandpass filter (LM358)
|   +-- HB100-ESP32.raw / .log         # Simulation output data
|
+-- Oscilloscope/                      # Hardware debug utility
|   +-- Oscilloscope.ino               # ESP32 Serial Plotter sketch -- ADC signal, DC bias, min/max/mean stats
|
+-- Screenshots/                       # Visual documentation
|   +-- block.png                      # System block diagram
|   +-- hardware.jpg                   # Hardware prototype photo
|   +-- ltspice.png                    # LTspice amplifier circuit schematic
|   +-- filter.png                     # Bandpass filter Bode plot
|   +-- matlab.jpg                     # MATLAB time-domain + FFT analysis
|   +-- adc_sampling.png              # ADC sampling determinism verification
|   +-- calib.png                      # Calibration & target acquisition graph
|   +-- dash.png                       # Dashboard vital sign telemetry
|   +-- magdis.png                     # Signal magnitude vs. distance plot
|   +-- streng.png                     # Barrier material degradation chart
|   +-- wavedrom.png                   # RTOS scheduling timeline
|
+-- LICENSE                            # MIT License
+-- README.md                          # This file

Getting Started

Prerequisites

Hardware

Component	Quantity	Notes
ESP32 Development Board	1	Any variant with ADC-capable GPIO 34
HB100 or CDM324 Radar Module	1	10.525 GHz or 24 GHz Doppler
Op-Amp (e.g., LM358 / MCP6002)	1–2	For active bandpass filter stages
Resistors & Capacitors	As per filter design	See LTspice schematic for values
Active Buzzer	1	3.3V compatible
LED	1	With appropriate current-limiting resistor

Software

Tool	Version
Arduino IDE	2.x+ (or PlatformIO)
ESP32 Board Package	Latest via Board Manager
`arduinoFFT` library	v2.0.0+ by Kosme
`WebSockets` library	By Markus Sattler (Links2004)

Installation

1. Clone the Repository

git clone https://github.com/chandansaipavanpadala/H.A.W.K-HumanActivityDetectionThroughWallsUsingMicrowaveKinetics.git
cd H.A.W.K-HumanActivityDetectionThroughWallsUsingMicrowaveKinetics

2. Hardware Wiring

Signal	Connect To	ESP32 GPIO
Bandpass Filter Output	`RADAR_ADC_PIN`	34
Active Buzzer (+)	`BUZZER_PIN`	25
Status LED (Anode)	`LED_PIN`	2

Tip: Use the LTspice/HB100-ESP32.asc schematic as a reference for the analog filter circuit design.

3. Configure WiFi Credentials

Open ESP32/comms_ui.h and update:

#define WIFI_SSID     "YourNetworkName"
#define WIFI_PASSWORD "YourPassword"

⚠ Security: Never commit real WiFi credentials to a public repository. Consider using environment variables or a .gitignored config file.

4. Compile and Flash

Open ESP32/ESP32.ino in Arduino IDE.
Select your ESP32 board model and COM port.
Install required libraries (arduinoFFT, WebSockets) via Library Manager.
Compile and upload the firmware.

5. Launch the Dashboard

After flashing, note the ESP32's IP address from your router's DHCP table.
Open Dashboard/index.html in any modern browser on the same LAN.
Enter the ESP32's IP address in the setup screen and click INITIALIZE UPLINK.

6. System Validation

On boot, the dashboard will show a cyan radar-sweep loading overlay while connecting.
Once connected, the amber calibration overlay appears for 15 seconds — stand clear of the sensor.
After calibration, the full tactical dashboard activates with live charts, BPM display, and range estimation.
Introduce a target subject in front of the radar — monitor the confidence gauge ramp from NO SIGNAL to CONFIRMED.
Upon confirmed detection: RED ALERT flashes, buzzer sounds, and the $HAWK,ALERT packet is broadcast.
Check the Extended Metrics row for real-time breathing BPM, heart BPM, estimated range, and detection duration.

Development Tools

Oscilloscope Utility

The Oscilloscope/Oscilloscope.ino sketch streams raw ADC values at 250 Hz to the Arduino Serial Plotter for hardware debugging:

Verify the DC bias from the op-amp circuit (expected ~1.65V from voltage divider; actual may vary)
Confirm the ADC signal is centered and not clipping at 0V or 3.3V
Validate the radar module is producing a clean IF signal
Automatic voltage statistics every 5 seconds: DC bias (mean), min, max, peak-to-peak swing, and clipping warnings

Tools -> Serial Plotter -> 115200 baud

LTspice Simulation

The LTspice/ directory contains the complete active bandpass filter schematic (HB100-ESP32.asc). Use LTspice to simulate and verify the filter response before building the physical circuit.

License

This project is licensed under the MIT License — see the LICENSE file for details.

Built for the Real-Time Operating Systems course — Amrita Vishwa Vidyapeetham

Project H.A.W.K. — Because every second counts when lives are behind walls.

Name		Name	Last commit message	Last commit date
Latest commit History 24 Commits
.github/workflows		.github/workflows
Dashboard		Dashboard
Diagnostic		Diagnostic
ESP32		ESP32
LTspice		LTspice
Oscilloscope		Oscilloscope
Screenshots		Screenshots
LICENSE		LICENSE
README.md		README.md

Folders and files

Latest commit

History

Repository files navigation

Project H.A.W.K.

Human Activity Detection Through Walls Using Microwave Kinetics

Table of Contents

Overview

Key Features

Key Applications

System Architecture

Hardware Architecture

Pin Map

FreeRTOS Task Architecture

Task Overview

Inter-Task Data Flow

Signal Processing Pipeline

Pipeline Stages

Key Design Decisions

Dynamic Auto-Calibration

Calibration Sequence

Non-Blocking Design

Why 75th Percentile Instead of Peak?

Distance Estimation

Empirical Validation

Limitations

Tactical LAN Dashboard

Dashboard Features

Tech Stack

Communication Protocol

$HAWK Protocol

Telemetry Fields

Screenshots & Visual Documentation

System Block Diagram

Hardware Prototype

LTspice Amplifier Circuit Schematic

Bandpass Filter Frequency Response

MATLAB Baseband Signal Analysis

ADC Sampling Determinism

Dynamic Auto-Calibration & Target Acquisition

Dashboard — Live Vital Sign Telemetry

Signal Magnitude vs. Distance (Through Drywall)

Vital Sign Degradation by Barrier Material

RTOS Scheduling Timeline

Repository Structure

Getting Started

Prerequisites

Installation

1. Clone the Repository

2. Hardware Wiring

3. Configure WiFi Credentials

4. Compile and Flash

5. Launch the Dashboard

6. System Validation

Development Tools

Oscilloscope Utility

LTspice Simulation

License

About

Topics

Resources

License

Uh oh!

Stars

Watchers

Forks

Contributors

Uh oh!

Languages