Android Developer – Offline Audio Event Detection Prototype

Project Overview

We are developing a rugged Android-based field prototype that detects short-duration mechanical audio events (1-2 second drill bursts) and logs one event per burst.

This is not a full application.

This is a focused prototype to validate detection reliability under real outdoor field conditions.

The goal is to detect specific short mechanical drill sounds and log them with high reliability while avoiding false positives.

Core Requirements

The system must:

Run fully offline on Android devices

Detect 1-2 second drill bursts

Log exactly one event per burst

Avoid double-counting during a single burst

Allow legitimate rapid second events (two bursts within 5-15 seconds)

Provide vibration feedback upon detection

Include large "Undo last event" and "Add event" buttons

Log timestamps locally and export to CSV

Allow configurable detection thresholds for field tuning

Field Conditions

Cold outdoor use

Intermittent wind

Device worn on chest

Consistent drill model and sound signature

Events spaced ~90 seconds apart on average

Occasionally 2-4 events within 10-20 seconds

Acceptance Targets (Field Test)

Detection rate ≥ 95%

False positives ≤ 1%

Double-trigger rate ≈ 0

Reliable across multiple Android hardware models

These metrics will be validated in real field conditions.

Technical Expectations

We are looking for someone with experience in:

Android development (Kotlin preferred)

Real-time audio processing using AudioRecord / low-level APIs

Digital Signal Processing (DSP)

Spectral analysis / frequency band detection

Event segmentation using silence thresholds

Designing debounce logic

Performance optimization for battery efficiency

A deterministic DSP approach is preferred for version 1.

We are open to leveraging well-maintained open-source libraries (MIT/Apache licensed) where appropriate.

If proposing open-source components, please specify:

Library/repository

License

Why it is suitable

How you would adapt it for short mechanical burst detection

Heavy ML models are not required unless justified.

Deliverables - Milestone 1

Android APK prototype

Real-time drill burst detection

Debounce logic to prevent double logging

Configurable detection parameters

Vibration feedback

Tap counter UI

Undo/Add buttons

Local event logging

CSV export

Source code in repository owned by client

Deliverables - Milestone 2

Field tuning based on real-world recordings

Threshold optimization

Reduction of false positives

Hardening for stability

Optional: Quote separately for adding GPS averaging functionality.

Timeline

Target timeline: 4-6 weeks total

Prototype must be ready for field testing within ~4 weeks.

Budget

Fixed price: $8,000-$10,000

Open to milestone-based structure.

To Apply

Please include:

Brief description of prior audio/event detection work.

Your proposed approach (DSP vs ML vs hybrid).

How you would prevent a 1.5 second drill burst from being counted twice.

Estimated timeline.

Estimated cost within budget range.

Generic proposals will not be considered.

Contract duration of 1 to 3 months. with 30 hours per week.

Mandatory skills: Android, Mobile App Design, Functional Prototype, Prototyping