A noise report arrives. An officer opens the recording, listens, skips forward, rewinds and listens again, trying to find the sound that prompted the complaint.

Multiply that process across hundreds or even thousands of submissions, and audio review can quickly become a significant drain of time.

The Noise App’s new audio classification feature helps teams get to the relevant parts faster.

Submitted recordings are automatically analysed to identify likely sound events. These events are then displayed within the web app audio player, allowing officers to jump directory between them instead of repeatedly searching through the full recording.

This means, less time manually navigating audio, faster initial triage and more time available to process investigations.

How audio classification works

There are no new steps for officers to complete. Classification takes place automatically when a reporter submits a recording though The Noise App.

The Noise App then:

• Analyses the submitted audio

• Detects likely sound events

• Matches them to the noise assigned to the report

• Displays relevant within the web app audio player

• Allows officers to move directly between detected sections

The classier uses noise source categories already supported by The Noise app, including barking dogs, construction noise and car alarms.

Officers remain in control of reviewing the recording, considering its relevance and deciding how the report should progress.

Helping teams manage growing report volumes

The Noise App processes more than one million noise reports every year.

Audio classification is designed to help environmental health teams manage this volume without review workload increasing at the same rate.

By making relevant sounds easier to locate, this feature can then help teams:

• Review submitted recordings faster

• Reduce repetitive listening and manual searching

• Triage cases more efficiently

• Identify potentially relevant evidence sooner

• Spend more time progressing and resolving complaints

Users can also provide feedback on classifications, helping improve the model’s performance over time.

What’s coming next?

We are developing additional safeguarding support to help identify recordings that may contain indications of harm, violence or threats for appropriate review.

This capability is still in development and will build on our work to make audio review more efficient, responsible and useful for investigating teams.