pensieve-speech

pensieve is an easy extensible server architecture to accept socket data such as sensor readings, images, etc. pensieve-speech is an extension of pensieve, with a handler for glass-based (or any) speech-to-text applications. As of time of writing, it is unstable and highly volatile. Do not expose to open flames or belligerent children.

Known Dependencies

• Python 2.7.x
• NumPy (pensieve doesn't need SciPy, just NumPy, but it doesn't hurt)
• OpenCV 2.4.x
• PyZMQ (optional, for streaming/pub-sub servers)
• ffmpeg (for converting audio files to flac format)

Installation (adapted from pensieve)

1. Clone:

   $ git clone [email protected]:rickyhopper/pensieve-speech.git

2. Install (after cd pensieve/):

   $ [sudo] python setup.py develop

   Note: This installs in development mode, which means Python modules are exposed directly from the source directory. You can then update your local copy to pull in changes from the remote repository, and/or make changes yourself. You can also use [sudo] python setup.py install for a typical installation (recommended: Python virtual environment).

3. Install ffmpeg.

Usage

Run server:

$ python -m pensieve.server

Run test client:

$ python -m pensieve.tests.client

Note: You can connect to a running server instance from any pensieve client.