Project ideas for audio machine learning

Smart projects for the Raspberry Pi Zero

Project ideas for audio machine learning

Project ideas for audio machine learning

Add a microphone hat to your Raspberry Pi and make some cool audio detection based projects. I assembled some ideas for projects if you want to dive into audio machine learning.

1. Ninja robot vacuum cleaner

Give your robot vacuum cleaner ninja-like stealth abilities. It could listen for human activity, like voice or footsteps, and hide when people. When nobody is around it can do its work without disturbing anyone.

2. Upcycle dumb devices

Most household machines have no WiFi to notify you when they are finished. Wouldn’t it be great to have a Pi Zero nearby notifying you about the progress of your washing machine, coffee machine or microwave?

3. Dancing robot

Make your robot recognize music and start dancing when it hears some. You can even discover some genres and modify the dance accordingly.

4. Crowd Monitoring

Listen to crows in airports, train stations etc. for events like screaming and gunshots. How about detecting a riot in a prison?

5. Animal monitoring

How often is and when is a bird feeding its children, when do they sleep? Which birds are in your area and how does population change over time? For researchers, this kind of data can be very valuable.

But also Farmers and Industry can profit from Machine learning. Monitoring the activity of Chickens, Fish, Cows, Horses, Sheeps, Bees and many more can detect panic, signs of diseases and the possible presence of predators.

6. Burglary Detection

Monitor your home for human activity. Let your home listen for glass breaking, footsteps and voices when you are on vacation and notify you when activity is detected.

7. Machine monitoring

Your ship engine is making some strange noise? The gearbox of a windmill is about to grind itself to death? Utilizing audio data might be able to avoid catastrophic failure and send a maintenance team before things go from bad to worse.

8. Traffic counter

How many cars are passing a nearby road, how many of them are trucks, when is the rush hour? Cars and trains make noise which means that traffic data can be captured from a distant place where power supply and network access is available.

9. Sleep health

Get some insight into how well you sleep. With audio recognition, you can detect if you’re talking or snoring during the night.

10. Geological monitor

Detect the sound of rockfall and avalanches or the noise of ice breaking from the Antarctic shelf. Combine this with weather data and you might even be able to forecast big environmental risks.

11. Voice activity detection

Listening through hours worth of recordings to find 10 seconds of speech is very costly and cumbersome. Using neural networks to find sections of speech can solve this problem. The same goes for recording and storing hours of audio data when only sections of speech are of interest.

12. Gender and age recognition:

Demographic groups behave differently. Being able to adapt to gender and age, businesses can better serve and target their customers. This can also be a good resource to capture statistics about demographic groups for example in all kind of areas.

13. Doorbell repeater:

Your doorbell is ringing or somebody is knocking on your door and you can’t hear it? Maby you aren’t even at home? Use audio detection to recognize all kind of rings and send you a notification to a device of your choice.

14. Voice controlled switch:

Control your lights, blinds, coffee-machine, and temperature with a simple voice command.

15. Speaker Verification:

Use the characteristics of your voice as a biometric password to secure access to your hobby garage.

16. Gather Higher Level Context:

If you have already some audio detection running you can extract higher-level information from it. You can set up a few smart speakers in your home/business and capture low-level information. This would be stuff like water boiling, door events etc.

Store the event timestamps and location in a database and label higher-level events like “kitchen is in use”, “everybody left the house”, “shifts are changing”. With this data, you can train a new neural net to detect these events.


Leave a Reply

Your email address will not be published. Required fields are marked *

%d bloggers like this: