Introducion:
First of all I want to appologize for bad english. I'm not a native speaker.

spokenRobot is a voice assistent, that combines the prefrences of two speech engines: pocketsphinx, which one giant prefrence is, that it runs offline, and the google speech api, which is very very precise at transcribing an audio file.
So spokenRobot uses pocketsphinx to just recognize one single keyword, 'computer' as default, and the google speech api, to recognize the following command.
If the keyword is recognized, a voice record starts, where you can call a command. This record will be sent to google, which returns its transcription.
This has the prefrence, that you dont have constant network traffic when you don't actually want Google to transcribe. eSpeak is used for voice output.
You can define your own commands in the commands-directory, but this will be explained later.

What you need:

• stable internet connection

• fully installed and useable microphone
  => you can test it by typing
  

    arecord "temp.wav" <br />
    	=> speak at microphone and cancel record with ctrl+c
    	
    aplay "temp.wav" <br />
    	=> plays the recorded file
  

  if you dont hear anything, your microphone is not properly working

Installation:
Just download the directory and put it somewhere in your home folder. then cd into it and type

 chmod a+x install.sh
 sudo ./install.sh

This will install espeak, sphinxbase and pocketsphinx next to a bunch of dependencies. You will be asked for a google speech api key. If you dont already have one, follow the instructions at

 http://nerdvittles.com/?page_id=21210 <br />

to create one. If everything worked well, you can start spokenRobot to run in background or test your microphone settings first.

Define your own commands:
To define your own command, you have to modify at least two files:

 commands/commandDomains.file	and
 commands/commands.file

In the first you have to define the so called 'Domain' of your command, which means some keywords, which have to be found in voice input, to actually start your command. So for example the command domain of the tell-time-command would be defined as follows:

 tell_time: [what, time], [says, time], time

The words in brackets are word combinations. This means, they have to be found both at least once in one voice input to start your command, so that e.g. you dont get the time by just saying "what" but by saying e.g. "what is the time?" This notation is a bit complicated, but you can find further information at commands/commandDomains.file

The second file contains the actual code, which will be executed, when your command has been spotted in voice input. The code has to be linux-bash code, since spokenRobot just copies this code into an .sh file and executes it. For more information of syntax, available functions, variables, etc. see commands/commands.file

If you add a command by modifying this two files, you dont need to restart any script, this commands will be automatically integrated.

There is one more file in the commands directory: audioList.file This file contains some output read out, if e.g. an error occured. You can change this file to e.g. translate its content into your language or to add some output for your own commands. The syntax and functionalities are however explained at commands/audioList.file

Change language:
In install.sh you are asked for an language code for googles speech api. You can find a full list of languages, supported by google at

 https://cloud.google.com/speech/docs/languages

To change it after installation, edit the file scripts/sendToGoogle.sh and change the position of "lang=en-us" in the link to your language-code.

The files in the command directory have to be translated in your language, and new commands can of course be defined in your language as well.

It is not recommended to change the language of pocketsphinx, since it is best at recognizing english, and since you just need it to recoginze a single word, it is not that hard to bear this limitation.

To change the voice output to a voice in your language, you have to modify the flag "-ven" of the espeak command in the
scripts/voice_output.sh
to the flag for your language, e.g. -vde for german. You can find a full list of supported languages at

 http://espeak.sourceforge.net/languages.html <br />

Alternatively you can download additional voices at

 http://tcts.fpms.ac.be/synthesis/mbrola.html <br />

Changing the keyword detected by pocketsphinx:
If you want for any reason to change the keyword that activates the recording for google speech api, to e.g. a name like "alexa" or "jasper", follow these steps:

 1. Create a file with some random words, containing the keyword
    you want pocketsphinx to recognize.
    This is the new word library and the less words are in it,
    the more it is likely pocketsphinx recognizes a single word in it.
    For example, if this file just contains your keyword, every
    input, even if it sound completely different, is recognized
    as this keyword. On the other hand, if it contains a few thousand
    words, it is very hard for pocketsphinx to recognize your keyword.
    Look at the keyword file in the directory, for an example of how
    to create such a file for the keyword computer.
    
 2. Go to the sphinx knowledge base generator site:
 
 	http://www.speech.cs.cmu.edu/tools/lmtool-new.html
 		
    and upload your file generated above.
    Then click "COMPILE KNOWLEDGE BASE" and download the .tgz file <br />
 
 3. Unpack it. <br />
 
 4. Put the files at 
 	
 	/usr/local/share/pocketsphinx/model/en-us <br />
 
 5. Replace the two occurs of the Number '1815' in scripts/autostart.sh 
    with the base name of your knowledge base. This will be a 
    four-digit-number, too.
    When you have replaced the two numbers, restart this script by
    typing:  <br />
 		killall pocketsphinx_continuous
 		scripts/autostart.sh & <br />
 
 6. Now the last thing to do, is, to change the "keyword" entry at 
    the config.conf file in the main directory from "computer"
    to your keyword.