forgive my poor English...

sound-processor

A small(gzip < 3KB), dependency-free audio signal processor, the flow is:

1. import original audio signal;
2. gaussian filter，make output more smooth;

when you getByteFrequencyData from AnalyserNode, webaudio has already apply a Blackman window, make up for gain loss, the parameters is unadjustable. But most of the time, the vision is prior to precision in audio visualization, so we apply a gaussian filter for signal smooth, the degree of smoothness is steerable.

3. weighting, include a-weighting and time-weighting. weighting simulate human subjective feeling of sound;
4. octave divide base on parameters given to SoundProcessor;

three demo（audio is load slow, please wait）：

• demo1；

• demo2；

• demo3；

install

npm install sound-processor

usage

import { SoundProcessor } from "sound-processor";

const processor = new SoundProcessor(options);
// in means original signal from getByteFrequencyData
const out = processor.process(in);

options

• filterParams: gaussian filter params，object，default to undefined，means do not filter：
  • sigma：gaussian distribution sigma，default to 1，means standardized normal distribution, bigger sigma means more smooth，usually between 0.1~250;
  • radius：filter radius, default to 2, bigger means more smooth；
• sampleRate：from audioContext.sampleRate, like 48000；
• fftSize：param for FFT，default to 1024；
• startFrequency：min frequency，default 0；
• endFrequency：max frequency，default to 10000，you can cut out any frequency section with startFrequency and endFrequency；
• outBandsQty：number of output bands, usually means number of visualization targets;
• tWeight：whether to apply time-weighting，default to false;
• aWeight：whether to apply time-weighting，default to true;

frequency cut

frequency of music is between 50~10000 Hz, but in practice, we can cut 100~7000 Hz. perfect frequency cut not exists.

contribution

clone this project，exchange content of package.json and package.dev.json, than npm run start.

license

MIT.