The Open Speech Recording project by Google is looking for volunteers from the Raspberry Pi community. Your voice can make AIY Projects better at responding to keywords.
The idea is to collect a lot of short audio clips of numbers and key words, such as "on" and "up".
The results will be used as an open source collection of voice and Google will use it to train voice models.
There are very few open source collections of speech data. Especially from areas around the world with varying dialects. We've already uploaded our voices and hope as many as you can help out.
In the end, the AIY Projects will be more responsive, and we'll have a set of open source voice keywords to use.
The Open Speech Recording project will capture a range of different voices saying key words, such as "one", "on", "yes", and "no".
You can help out using any computer with a microphone or your AIY Projects Voice Kit. Open the Chrome/Chromium web browser and perform the following.
- Visit aiyprojects.withgoogle.com/open_speech_recording.
- Read the terms and conditions and click I Agree.
- Click Allow in the "Use your microphone" alert popup.
- Click Record
- Read out each word as it appears on the screen. After you've read the word, it will move automatically to the next one. We had 110 words in total. It takes about 5 minutes.
- At the end of the list, an alert will appear with "Are you ready to upload your words?". Click OK, and the words will begin uploading. (Click the blue Upload button if you don't see "Uploading".)
Open Speech Recording: open source voice data
The goal is to capture crowd-sourced speech clips to be used as training data for TensorFlow voice recognition models. Google intends to capture the speech then train the model.
Google informs us that both the model and training data will be open source and made available on the AIY Projects website.
According to the AIY Projects website:
Both the model and training data will be open source on the website once we've prepared it all. There are very few open source collections of speech data where individual words are spoken by a large number of different people. This makes it tough to build good open source examples of detecting spoken keywords such as "Yes", "No", "On", or "Off". To fill this gap, the AIY team is hoping to gather single spoken words from several hundred people and then release the resulting data set under an open license, together with an example of how to use it to create simple speech command classifiers.
Click here to learn more and help out with Open Speech Recording.
We hope to have more AIY Projects for voice kits available soon. Sign up for our newsletter to quickly find out when the new kits are available.
