This master’s thesis presents a real-time sound source localization system
based on a Raspberry Pi with a six-channel Seeed ReSpeaker microphone
array. The system is designed to detect and determine the direction of
branch-breaking events in forest environments, where the device is mounted
on a processor or harvester. Localization is based on the time difference
of arrival (TDOA) between selected microphone pairs, enabling estimation
of the direction of the incoming sound. The implementation is written in
Python and uses a multithreaded architecture to record and process audio
signals simultaneously. Frequency-domain analysis, energy thresholds, and
kurtosis-based filtering are applied to distinguish relevant branch sounds from
background noise and engine activity. The results confirm the feasibility
of a low-cost embedded solution for directional sound detection to support
automation in forestry.
|
