Animal Watch: Wildlife Acoustic Monitoring Node

Aalto University (Technology Entrepreneurship) · Finland · 2024 · Academic project · Group project (Group 4)

A solar-powered Raspberry Pi sensor node prototype for reducing elk-vehicle collisions: microphone arrays for direction-of-arrival estimation, an ML sound classifier, and environmental sensing, developed as a technology entrepreneurship project at Aalto.

Raspberry Pi
Audio ML
YAMNet
Sensor node
Entrepreneurship

Overview

Elk-vehicle collisions are a real safety problem in the Nordics. Animal Watch is a monitoring node that listens for elk vocalizations, estimates where the sound came from, and reports to a base station so warnings can reach drivers. The project covered the system design, a working solar-powered prototype (see photos), a business case, and regulatory considerations (CE marking, ISO 9001, and relevant EU directives).

Technical approach

Raspberry Pi node with two I2S MEMS microphones for direction-of-arrival estimation, a BME280 temperature/humidity/pressure sensor, and a TFT status display
Sound classification with a modified YAMNet model trained on augmented elk-vocalization data
Solar-powered enclosure with base-station reporting
Idea-pitch video and commercialization plan alongside the technical report

Media

Black enclosure with a small solar panel mounted on top — The assembled sensor node prototype with its solar panel.

System diagram of the Animal Watch technology stack — Technology stack: sensing (microphones, environment sensors), control, communication to the base station, and the solar power unit.

Wiring diagram of Raspberry Pi, microphones, sensor, and display — Wiring plan: Raspberry Pi with two I2S MEMS microphones, a BME280 environment sensor, and a TFT display.

Raspberry Pi and breadboard during prototyping — Breadboard bring-up of the node during development.

Open prototype enclosure showing wiring — Inside the enclosure during assembly.

Also listed under: Software & Machine Learning, MEMS & Microfabrication