협소 공간 내에 생존자 탐색을 위한 인명 탐지 모듈과 최소 제곱법 기반 음원 위치 추정 알고리즘에 대한 연구
A Study on the Survivor Detection Module and Least-Squares Sound Source Localization Algorithm for Victim Search in Narrow Spaces
1
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
In this study, we designed a survivor detection module for victim search in narrow spaces and proposed Least Squares-Time Delay of Arrival (LS-TDOA) based sound source localization and visualization algorithms to be integrated into the module. The module was designed as a compact structure capable of navigating through confined spaces, with an expandable umbrella-like mechanism for size adjustment when needed. Additionally, the module is equipped with a gas sensor for monitoring the surrounding environment and a thermal camera for detection in low-light conditions, alongside a microphone for sound source detection. To improve the sound source localization algorithm, we utilize voice activity detection (VAD) to distinguish human voices, thereby improving robustness to environment noise. The least squares method was employed to calculate the coordinates of the sound source, minimizing localization errors. Finally, the performance of the proposed algorithms was verified through simulations across various environments, and field experiments confirmed the effectiveness of the sound source visualization and localization.
Abstract
In this study, we designed a survivor detection module for victim search in narrow spaces and proposed Least Squares-Time Delay of Arrival (LS-TDOA) based sound source localization and visualization algorithms to be integrated into the module. The module was designed as a compact structure capable of navigating through confined spaces, with an expandable umbrella-like mechanism for size adjustment when needed. Additionally, the module is equipped with a gas sensor for monitoring the surrounding environment and a thermal camera for detection in low-light conditions, alongside a microphone for sound source detection. To improve the sound source localization algorithm, we utilize voice activity detection (VAD) to distinguish human voices, thereby improving robustness to environment noise. The least squares method was employed to calculate the coordinates of the sound source, minimizing localization errors. Finally, the performance of the proposed algorithms was verified through simulations across various environments, and field experiments confirmed the effectiveness of the sound source visualization and localization.