== Introduction ==

This software tries to implement (currently undocumented) Google Speech API, using CMU Sphinx's pocketsphinx gstreamer plugin as the recognizer backend.

Google Speech API is discussed here: http://mikepultz.com/2011/03/accessing-google-speech-api-chrome-11/

== Prerequisites ==

Install pocketsphinx, with its gstreamer plugin.

Install Ruby, Sinatra, .. [TODO]

Create out/ directory for logging request data:

mkdir out

== Starting server ==

Add local lib directory (where CMU Sphinx is probably installed) to GST search path:

export GST_PLUGIN_PATH=/usr/local/lib

Start server:

ruby speech-api.rb

== Using server ==

Create test.flac file in 16kHz, mono

Send it to server:

curl -v -X POST --data-binary @test.flac -H "Content-Type: audio/x-flac; rate=16000"  http://localhost:4567/recognize