soundarray

Active Development

← Back to Signal Processing


Concept

The ability to capture, localize, and classify complex soundscapes on edge devices or via remote streaming, providing structured insights to an agent framework.

Quick Facts

   
Status Active
Language N/A
Started 2026

What This Is

A spatial audio processing system using Raspberry Pi and microphone arrays. It combines sound source localization (Time of Arrival, beamforming) with ML-based classification (vehicles, wildlife) using ODAS for DSP and YAMNet for edge inference, publishing structured detections to an agent framework via MQTT.

Key Features

Processing Pipeline

Mic Array (8-ch PCM via ALSA)
     FFT
GCC-PHAT (Localization)
     Azimuth/Elevation
Beamforming (Source Separation)
     Mono per source
Mel Spectrogram (librosa)
    
YAMNet TFLite Inference
    
JSON/MQTT  Agent Framework

Roadmap

Tech Stack

C++ (ODAS), Python (librosa, NumPy, PyAudio), TensorFlow Lite (YAMNet), MQTT (Mosquitto), Raspberry Pi


← Back to Signal Processing