[PYTHON] I talked to Raspberry Pi

Overview

• The author is a beginner in Python, so the purpose is a brief introduction. This is my first post, so I look forward to working with you.

Jasper is written in Python, An open source voice manipulation application written by two Princeton University students. Probably because development started last year, there is still little contributing, so If you can write Python, please do.

Because there are many compatible STT (voice → text) and TTS (text → voice) engines, It's quite fun. Currently, it is possible to communicate with Twitter and Evernote via API.

For example (all translated into Japanese): You "Jasper!" Jasper "Pip" You "tweet" Jasper "What are you tweeting?" "Pip" You "I hope it will be fine tomorrow" Jasper "Tweet"

In short, it is a certain butt that moves with Pi.

Installation procedure

1. Open Official Documentation
2. As written, just hit apt-get etc. to get the dependent file (I added it later because there was a problem)
3. Speak enthusiastically in English while watching the recognized text flow as a log

Precautions at the time of introduction

The author introduced it using Method3 (Rasbian already installed), I had some problems.

First of all, when introducing PocketSphinx, which is inevitable no matter which method is used, I have a lot of dependent files, but it takes a lot of time to build OpenFST. It will take several hours. Let's wait patiently.

Also, it seems that it is the version of OpenFST (1.3.3) described in the document. Because it is moss due to an unidentifiable cause (sorry) I tried it in 1.4.1 and it was okay.

And with python populate.py, Jasper is made to" remember "various personal information. When specifying the address (location) with the name of the nearest large city, even if you type "Tokyo" etc., it will not be recognized. Apparently, it only supports American cities, and the author does not use it.

STT edition

The author seems to be aware of the default Pocket Sphinx, so It uses the API of Google Speech. The setup method is described in the document, so I will omit it a lot, but https://developers.google.com/ Get API_KEY from and write it in profile.yml.

• However, there is a limit of 100 times a day (?), So be careful when verifying.

Miscellaneous feelings

I think that if there is an SST that supports Japanese, it will be exciting in Japan as well.

Even so, it's fun to be able to move various things with voice recognition on such a small machine.

The author introduced Cygwin on Windows and Ssh to Windows through a module I wrote and I play games such as launching Steam games on my PC from the sofa.

Postscript (January 27, 2015)

I'm currently trying to control the air conditioner. If successful, I would like to write an article using photos.