This is a prebuilt dataset for a keyword spotting system based on a subset of data in the Google Speech Commands Dataset, with added noise from the Microsoft Scalable Noisy Speech Dataset. It contains 25 minutes of data per class, split up in 1 second windows, sampled at 16,000Hz. The dataset contains:
- Yes - one second samples with only the word "yes" in it.
- No - one second samples with only the word "no" in it.
- Unknown - one second samples of other words.
- Noise - one second samples of background or static noise.
Importing using the studio
Go to Data acquisition and click on the 'Upload icon'. Follow the instructions on the screen.
Or, import with the CLI
Open a terminal or command prompt, and navigate to the place where you extracted the file. Then run:
$ edge-impulse-uploader --clean
$ edge-impulse-uploader --label noise --category split noise/*.wav
$ edge-impulse-uploader --label unknown --category split unknown/*.wav
$ edge-impulse-uploader --label no --category split no/*.wav
$ edge-impulse-uploader --label yes --category split yes/*.wav