Indoor Acoustic Occupancy Dataset: Smartphone-Based FMCW Sonar in Diverse Room Geometries

This dataset contains active acoustic sensing data collected for the purpose of indoor occupancy estimation and human presence detection. The data was acquired using a commodity smartphone (Samsung Galaxy Z Flip6) operating in a monostatic sonar configuration. It is specifically designed to benchmark data-driven and Physics-Informed Neural Networks (PINNs) in challenging acoustic environments. Experimental Setup The dataset covers three distinct room geometries chosen to maximize variance in reverberation and scattering properties: Room 1 (Standard Living Environment): ~16 m², moderate absorption (furnished). Room 2 (Reverberant Chamber): ~12 m², sparse furnishing, hard surfaces. Room 3 (Confined Space): ~8 m², high clutter density, short mean free path. Signal Characteristics Signal Type: Linear Frequency Modulated Continuous Wave (FMCW) chirp. Frequency Range: 17 kHz – 22 kHz (Near-Ultrasound). Bandwidth: 5 kHz. Chirp Duration: 20 ms. Sampling Rate: 48 kHz. Hardware: Samsung Galaxy Z Flip6 (acting as both Transmitter and Receiver). Data Content & Protocol The dataset includes recordings for two primary states: Empty (Baseline): The static acoustic signature of the unoccupied room. Occupied (1 Person): Recordings of a single human subject. The protocol includes: Static NLOS: Subject located in Non-Line-of-Sight zones (corners, behind furniture). Micro-Motion: Natural breathing and slight postural shifts to induce micro-Doppler variance. Data Format The data is provided as processed Normalized Cross-Correlation (NCC) profiles (Pulse Compressed Echo Profiles). File Format: .csv (Comma Separated Values). Columns: Time-domain samples of the cross-correlation function. Resolution: High-resolution echo profiles suitable for Deep Learning input. Potential Applications Indoor crowd counting and occupancy detection. Physics-Informed Machine Learning (PINNs) for acoustics. Low-power edge sensing and smart home automation. Analysis of multipath propagation and human acoustic absorption. 3. Keywords Acoustic Sensing, FMCW Sonar, Occupancy Detection, Indoor Localization, Physics-Informed Neural Networks, Smart Home, Smartphone Sensors, Active Sonar, Human Presence Detection, Deep Learning 4. Creators / Authors Family Name, Given Name: Altinses, Diyar ORCID: https://orcid.org/0009-0005-7928-5874 5. License Creative Commons Attribution 4.0 International (CC-BY 4.0) 6. Version 1.0.0