Assistive Technology -Speak To Me Speech/Voice Recognition

    Speech recognition was how I started in the field, back in 1983. At that time, all that was available was discrete (one 'utterance' at a time, must pause between utterances), speaker-dependent (user must provide a consistent sample of each utterance) recognition. And it actually worked pretty well, even with the old Apple II's. In fact, this is how Dragon Dictate (to a great extent) worked. Unfortunately, Dragon Dictate was discontinued once Dragon was purchased by ScanSoft.com. The big selling SR package these days is NaturallySpeaking, a fine program for those that speak in 'typical' and continuous speech. And those are *not* the types of users I typically run across.

    My populations actually do better speaking the 'old' way, discretely. And their speech is so different from 'typical' that they get very poor recognition rates from SR engines that are looking for typical speakers.

    Now, a great advantage that Dragon had over what I'm proposing, was the inclusion of a 40,000 word vocabulary, already set up. And it even had pre-trained samples stored, of different types of users. As one used the product, it learned the user's voice even more and accurary increased.

    So I've made a nice little program *for Windows only* that works with discrete utterances and is speaker-dependent. There is *no* vocabulary already set up. I see three useful functions:

1) Command: All commands that would normally require a mouse or keystrokes, can be accomplished by speaking. When I say "speaking", because the software is speaker-dependent, it requires someone, or some *thing* to provide samples of how he/she/it is going to say that utterance. Notice I include "thing" and "it" because they can be an Augmentative Communication device. Yes, that works *great* because it is so consistent.
     So here are some starting points: Open, Save, Launch a program, Close, start the cursor moving up/down/left/right, Faster, Slower, Stop, Click, Date, etc. Anything that needs be done by mouse or keyboard (or any combination), for *any* software, can be set up within SpeakIt.

2) Control:    Another example would be a kids CD, like Reader Rabbit Toddler. You can either control it with the mouse movement commands, or you can setup a complete 'macro' to move to a certain location and click. I, personally, find that type of setup too time consuming (similar to the ClickIt requirements of creating hot spots), but it will have its uses.

3) Limited dictation: As I state above, Dragon Dictate was the 'king' for dictation for non-typical speakers, while Naturally Speaking is the 'king' for typical speakers. The former had extensive vocabulary built-in, and the latter has actual Artificial Intelligence built into it. SpeakIt has neither. But because *you* set up the vocabulary, and have the user train it, you can set up words as well as command/control. That is, the utterance "is" will yield the keystrokes "i" and "s", within any software. Or, the utterance "heading" could yield the user's name, address, date, and "Dear".
     I have several testers that *are* interested in creating these word vocabularies and would love to have them available to SpeakIt users someday. Other than the work involved in creating the vocabulary, the down side is that each utterance must be trained individually. I believe Dragon Dictate had a training feature whereby one could speak an entire sentence and in doing so, train a bunch of words at the same time.

The version on this page expires after 30 days of first use. I hope you will provide feedback and suggestions to me.

(Read my Downloading notes to learn more about
downloading, decoding, and decompressing)

2-24-06 - For those of you getting errors running the Sound System Setup Wizard, that utility only serves to set the input microphone volume/sensitivity. You can set that manually when the above error happens. Usually, though, it's not necessary.

Do a Training of a Voice Command, and if the waveform is in the middle of the black area, you're good. If the waveform 'clips' at the top of its range, then the input volume needs to be decreased. If the waveform is almost non-existent, then the volume must be raised.

To change the microphone volume:
1) Double click on volume speaker icon near time in taskbar.
2) Click Options in menu bar, then Properties.
3) Click on Recording then OK.
4) Adjust your Microphone accordingly.

Speak To Me! Software for Download 1) Download the software. It will let you try it for 30 days. You can then... 2) Purchase a Registration Code (or you can purchase the Code at any time)
SW-28R - Registration Code for Speak To Me! for ALL Windows - $59	Download Now (3.3 MB) or Purchase

Speak To Me! Software for Shipment
SW-28 - Speak To Me! for all Windows...$59	Order