Design and Evaluation of Multi-Node Acoustic Sensing Systems for Indoor and Outdoor Environments

Date of Award

8-15-2026

Degree Name

M.S. in Electrical Engineering

Department

Department of Electrical and Computer Engineering

Advisor/Chair

Hui Wang

Abstract

This thesis presents a distributed acoustic sensing and time-difference-of-arrival (TDOA) localization framework that progresses from a single acoustic receiver node to fixed indoor multi-node localization and GNSS-assisted outdoor deployment. The system combines microphone acquisition, pulse-per-second (PPS)-referenced waveform reconstruction, wireless audio transmission, receivergeometry preparation, pairwise generalized cross-correlation with phase transform (GCC-PHAT) delay estimation, and nonlinear source-position estimation. A single-node sensing platform was developed using an INMP441 digital microphone, an ESP32- WROVER-E controller, an external PPS timing input, and User Datagram Protocol (UDP) audio transmission. The host reconstructs waveform records from frame-order and node-level timing metadata rather than using UDP packet-arrival time as the acoustic time axis. This receiver-node design was extended to a fixed five-node indoor localization system with a shared PPS timingdistribution subsystem and measured two-dimensional receiver geometry. Indoor evaluation showed that localization performance depended on source characteristics and pairwise delay quality. A broadband hand-clap event retained all ten receiver-node pairs and produced a localization error of approximately 0.068 m. In contrast, a low-frequency-dominant JBL playback condition retained two pairs and produced an error of approximately 0.290 m. A systematic frequency–duration evaluation identified an empirically observed region of increased error and variability between approximately 8 kHz and 12 kHz. A real-time extension using frameindex-aligned waveform grouping, geometry-constrained delay screening, rank-based pair selection, and robust optimization was evaluated across 80 trials. Five-node configurations produced more compact and near-reference estimate distributions than corresponding four-node configurations under the evaluated indoor condition. The framework was further extended to an outdoor GNSS-assisted rover system. Local GNSS PPS references provided timing anchors, while RTK-derived rover positions were transformed into a local East–North coordinate frame and corrected to acoustic reference locations. A preliminary outdoor evaluation of twelve field recordings produced 216 correlated analysis-window configurations. Across all configurations, the median horizontal localization error was 0.0128 m, and 64.4% of configurations produced an error below 0.05 m. The results establish a practical foundation for future systematic outdoor evaluation, real-time field deployment, and multi-sensor acoustic–thermal integration.

Keywords

Acoustics, Electrical Engineering

Rights Statement

Copyright 2026, author

Link to Full Text

Share

COinS