Browser Signal Processing

The browser-signal-processing library offers some signal processing functions designed for speech applications. Some of them are executed directly in the Javascript engine, while others exploit remote web services.

The functions executed directly in the browser include both simple algorithms that fit into a few lines, such as RMS, and advanced algorithms such as PSOLA. The supported web services cover some very complex signal processing such as automatic phoneme segmentation (WebMAUS).

The library also includes some helper functions to read and write specific on-disk formats (such as WAVE or TextGrid) and to simplify the use of such browser APIs as WebAudio or XMLHttpRequest.

Usage Hints

The signal processing functions either operate on an array of numbers (number[]; the most basic form of a Javascript digital signal), or on a Float32Array, or on an AudioBuffer object. Some functions need to know the signal's sample rate. It needs to be passed explicitly unless the signal is passed as an AudioBuffer object, which has the sample rate integrated.

Functions that need to exchange data in terms of some on-disk format (generally functions that call remote web services) accept these data in two ways. They can either be already encoded in said format and be passed as ArrayBuffer objects (which are easily retrieved via XMLHttpRequests), or as Blob objects. Or they can be in an unencoded form: string for plain text files, AudioBuffer/Float32Array/number[] for audio files.

Functions that call remote web services either connect to them via WebSockets or use asynchronous XMLHttpRequests to call RESTful APIs. The results are made available to the calling Javascript/Typescript code by means of Promises that resolve to the result values as soon as they are available.

Usage Examples

@todo add usage examples

License and Authors

The libary is published under the terms of the MIT license (see file LICENSE).

The main author is Markus Jochim ([email protected]).

Some code is re-used from other projects, most notably the EMU-WebApp. Contributors are:

• Raphael Winkelmann
• Georg Raess
• Matt Diamond
• Jam3