Announcement

Collapse
No announcement yet.

Voice Commands -- AI -- Samantha

Collapse
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • Voice Commands -- AI -- Samantha

    Voice commands are a part of this program that I neglected for a long time. The code was there, but nothing was built around it, aside from a few commands for demonstration purposes. And I eventually broke those one day and never even noticed. However, voice commands are now a bigger part of DriveLine. From early on, DriveLine could read text out loud. Combining voice input and voice output brings us a concept that I'm calling "Auditory Interaction". See what I did there?

    There will be three parts to this explanation.

    Part 1 is the concept.

    Part 2 is the list of commands and responses.

    Part 3 is the technical "how it works" explanation.


    Part 1.
    I would like to introduce you to Samantha. She's inside your CarPC. She knows everything that DriveLine knows, and then a little bit more thanks to the internet. Depending on the devices you have connected to your PC, she knows where you are, where you might be going, how long it's been since you shut down or started up the PC, the time of day, the weather, your name, her name, the files in your playlist, the temperature of your engine, and other things. She can perform simple tasks, and answer some questions. A lot more can (and will) be added to her functionality, but this is the start.


    Part 2.
    This is the list of commands. If Samantha's code is all working correctly, then none of the commands need to be spoken exactly. For example, you could say "show playlist" or "show me the playlist" or "show the playlist screen" and any of those would result in Samantha displaying the Playlist screen. The idea is that you can speak to her the way you would speak to a normal person. You don't need to speak like a robot. And neither should she. Here's a basic list of the commands/keywords she understands and some explanations. This list will continue to expand.

    First, the simple commands that don't really do anything.
    "What time is it"
    "What day is it"
    "Where am I", "Where are we"
    "Who are you", "Who am I"
    "How are you", "How do you feel"



    More complex commands:

    Eject - This tells Samantha to eject a drive.
    examples: "Eject the D drive", "Eject drive D", "Eject D"

    Directions / Navigate - This command will tell Samantha to retrieve directions from the internet to a particular location.
    examples: "Navigate home", "Directions to work"

    Play - This will tell Samantha to play a different song. This is probably the most complex command.
    examples (Media): "Play next song by Green Day", "Play previous favorite", "Play random song", "Play random Green Day song", "Play random favorite"
    examples (Radio): ----- I still need to program this.

    Switch - This switches media sources.
    examples: "Switch to dab radio", "Switch to media", "Switch to current playlist"

    Dim / Darken - This will decrease the brightness of the screen.
    examples: "dim the screen", "darken the display", "dim the monitor"

    Undim / Brighten - This will increase the brightness of the screen.
    examples: "undim the display", "brighten the screen"

    Weather / Forecast - Samantha will tell you about the weather.
    exmaples: "how is the weather", "what is the forecast for Tuesday"

    Show - This tells the skin to display a different screen.
    examples: "show the playlist", "show the map"

    Repeat - Samantha will repeat the last thing she said.
    examples: "Repeat that", "Repeat the last thing you said"


    There are a few more commands that I am testing, and I probably need suggestions for other commands to add.

    More info to come..
    Last edited by ClockWorK; 09-03-2014, 09:49 PM.

  • #2
    Question "simple"....
    ONLY works and always speaking in English or the language of the OS \ CarPC ?

    Comment


    • #3
      Question "simple"....
      ONLY works and always speaking in English or the language of the OS \ CarPC ?

      Edit:
      I mean the questions.

      Comment


      • #4
        Unfortunately, the voice commands probably only work in English at the moment. However, that could change in the future, but I might need some help from someone using a non-english language for testing.

        Comment


        • #5
          Perhaps making a list of "fixed" commands with dictionary file translated into various languages​​?
          Maybe not this possible?
          I ask ...

          Comment


          • #6
            I am having good luck with the voice commands. I really like the AI too. I do not know how to change the voice, I have windows 7 and only 1 voice is shown in the drop down list. Anybody know where to get more voices? I like the female voice I have now, but a nastier female voice might be more pleasing!

            Comment


            • #7
              I was considering buying a voice from the link below, just to try it. Their prices are low enough that it's not a huge investment.
              https://www.cereproc.com/en/storesapi

              Try googling for SAPI voices and maybe you can find a free one to try.

              Comment


              • #8
                I tried all the sample sounds, and Microsoft Anna actually sounds better than their choices. I was thinking maybe one of the British female voices might be ok, but what I really want is a voice like Brittany Murphy's. I gotta get in touch with my inner Skeeze!

                Comment


                • #9
                  I totally understand what you want. Unfortunately, I cannot program emotion into the voices -- only words. But yeah, I wish Anna sounded a little sexier or more playful or something.

                  Comment


                  • #10
                    NeoJulie

                    NeoSpeech - Text to Speech US (TTS)


                    Try this,...

                    http://www.neospeech.com/

                    I use US English "Julie" and I'm very happy with the results.

                    Also, slowing down the voice a bit in tts windows settings gives it just a little more of that sexy appeal edge.

                    As far as more playful? Hmm.

                    Enjoy.
                    Last edited by RAWPWR; 06-19-2014, 09:16 AM.

                    Comment


                    • #11
                      Originally posted by ClockWorK View Post
                      Unfortunately, the voice commands probably only work in English at the moment. However, that could change in the future, but I might need some help from someone using a non-english language for testing.
                      As far i know the only international support (for win os) is Microsoft Speech Platform already used on some speech recognition tools, but it might be not supported on Win8
                      (i'm using it on win7)

                      Comment


                      • #12
                        I made a Video of my latest carcp progress, and it shows some of the voice commands working in Driveline:

                        Skip to 1:30 into the video to see just the voice commands in action...




                        As far as other voice commands to add, I think would help are:

                        "Switch Source"... Because Switch HD only works 1 in 10 tries. I spoke that command into windows speech to text to see what it is hearing, and there are too many "pitch as thee" and other things interpreted.

                        Also, would it be hard to add a speech to text box that pops up when we press the talk button, that we can have show what the speech recognition hears, as it would help in training us to speak our commands more successfully.

                        Another feature that might be cool is if we could save multiple playlists, and have saved favorites in each. Then a "Switch Playlist" command would really be cool.
                        Last edited by rray; 06-21-2014, 05:20 PM.

                        Comment


                        • #13
                          Originally posted by rray View Post
                          I made a Video of my latest carcp progress, and it shows some of the voice commands working in Driveline:

                          Another feature that might be cool is if we could save multiple playlists, and have saved favorites in each. Then a "Switch Playlist" command would really be cool.
                          You can save playlists. And the Favorites are global. So if you mark a song as a favorite, and then later open a playlist with that song in it, it will still be marked as a favorite.

                          Try "Open playlist" followed by the name of the playlist. DriveLine will search the "My Playlists" folder for a match.

                          I'm not in the best frame of mind at the moment. Just broke up, had a few drinks, about to smoke my feelings away for the night.. But hey, DriveLine always gets better when I have nothing else distracting me.

                          EDIT: Also, thanks for the plug. It's awesome to see DriveLine being used to its full potential.

                          Comment


                          • #14
                            rray i am speechless by ur project ....! very nice setup.
                            Vw Sharan 2007 1.9tdi || Exide 100Ah
                            Fuzzy 945gme + core duo T2400 + 2gb ddr 667 + pci nvidia 5500 256mb + kingston v300 60gb + hdd 2.5" 100gb 5400 + Sure Gps module
                            Primary monitor 8" 800x600 || Secondary monitor 15.4" 1280x800
                            Powered by M2 atx

                            Comment


                            • #15
                              I found that as long as the windows are rolled up, the speech recognition is working great, even when I have the air conditioner fan up full blast. I am really liking voicehose tie commands and the AI's turn by turn directions in the newest test version.

                              I would like to request one more feature if possible, would like to have a voice training screen that shows what the voice recognition hears for times she don't understand what I am saying. Like a practice screen that echo's back the spoken command in text form with a large font for easy reading.

                              I am thinking a voice training button that opens a screen that has the Voice Command button and a text window to display what is heard, so we can teach our voice to speak correctly. I have trouble switching to some playlists, one named BOM and one named DNC are difficult to get to come up through voice commands. Maybe new commands "Next Playlist", "Previous Playlist", and "Random Playlist", in addition to of "Switch Playlist XXXX"

                              Comment

                              Working...
                              X