About

This tool transforms wav files into spectrograms. Either as a PNG or as a json for further use in tensorflow.

How to use:

1. git clone https://github.com/360disrupt/nodejs-spectrogram.git && cd nodejs-spectrogram
2. Rename .env.example, change the default config to your demands and put the audio files to convert into the AUDIO_IN folder.
3. npm run dev (Maybe adjust ram settings in the package.json)

You need to follow the instructions of the package canvas to draw spectrograms.

Envs

• AUDIO_IN_FOLDER: {string} the folder to look for audio files
• WATCH: {boolean}, watch the audio in folder for new files
• OUT_FOLDER_JSON: {string} the folder for the json spectrogram
• OUT_FOLDER_DRAW: {string} the folder for the png spectrogram
• GENERATE: {boolean}, use a dummy sine wav (for testing purposes)
• DRAW: {boolean} create a png
• SEND: {boolean} send spectrograms via http
• SEND_URL: {string} url to send spectrograms to
• RESOLUTION: {number, root of 2} how many samples are in one sequence
• CROP: {boolean} crop files to the smallest file to have the same sequence length
• CROP_MAX: {number} crop files to a max of sequences
• DEBUG: {string} show debug logs

Misc

Let me know if you need this as an NPM package. I can convert it and make the process envs to an options object.

Useful resources

About fft/spectrometer in python https://www.oreilly.com/library/view/elegant-scipy/9781491922927/ch04.html

Digital signal processing https://dsp.stackexchange.com/

License

MIT, see license file