MAKING SPEECH NATURAL
AND
TALKING YOUR WAY NATURALLY

MAKING SPEECH NATURAL
Synthesis And Synthesizers

Adventures In The Quest For Natural Sounding Synthesized Speech.
By definition, synthesized speech is "synthetic" sound. This is"electro - mechanical", and not natural. In reality it is "sound" that is synthesized to resemble speech. Just as sound can be synthesized to resemble musical instruments. Accomplished by mathematic programming because recorded voice WAV Based Systems take up disk storage space and processor power.
Synthesis To Sound.
To the human ear and expectations music instruments are simple wave forms compared to the human voice. Music synthesis can synthesize the generic sound of a music instrument. Music synthesizers still can only make an approximation to reproduce the individual sounds of a specific music instrument ; or a specific musician. In speech the complication is with the human voice which is both instrument and musician.
Synthesis To Speech.
The human voice is complex and further complicated by the human factors of individual preference and perception. Even before pronunciation and enunciation, there are numerous factors, and sub - factors, that go into what we "perceive" as making speech natural.
For example even with today's imperfections in synthesized speech:

1. Some users prefer hearing the synthesized female voice, rather than the synthesized male voice.
2. Some children prefer hearing a synthesized child's voice.
3. Just as some senior adults prefer hearing the synthesized voice of a mature adult.
4. Then preference is further complicated by National and Regional pronunciations and common usage.

As computerized speech improves, the size of the mathematic programs grow. Short of a mainframe, PC computer speech will continue to improve as processor power and storage space increase. If not evolving into a WAV Based System; then eventually evolving into a WAV Quality System.
For now, like it or not, there is no Natural Sounding Text To Speech available for PCs. If you want to improve the currently imperfect text -to- speech applications your choices are:

1. Manually change the phonetic pronunciation vocabulary.
or:
2. Change your expectations to accept the program's pronunciation vocabulary.
Out of the box, most users elect to change themselves, rather than the program.

TALKING YOUR WAY NATURALLY
Through Command And Control
As Well As:
Recognition And Dictation

Here technology takes two separate but related parallel paths. Before coming together again in one application:
1. Command and Control speech utilities allow the user to voice input verbal commands to control the computer and its' application programs.
2. Recognition and Diction speech programs allow the user to voice input text to the computer's application programs.
3. Between the two are the speech programs which allow the user to voice input verbal commands to control the computer and its' applications. As well as allow the user to voice input text to the computer's application programs.
Speech Dictation programs have only recently made the transition from "special" to finally reach the marketplace as a general application. This is an important event as special programs and their special prices did not have the advantage of mass marketing and the price benefits of scale. However, be aware that your individual special needs and expectations may differ from the manufacturer's specifications; as not all programs are created equal.
To Train Or Not To Train.

In Computer Speech Command and Control, Recognition and Dictation, both you and/or your computer program will need to "adjust or train". Even the newest out of the box "self training" programs require user input.
Self training systems require user input to enhance the program's learning process. Adding text vocabulary as well as spell checking and correcting are functions which enhance greater program accuracy.
Out of the box only the simplest single purpose programs with limited and preset vocabularies require no further training.
1. Either the user trains the program to recognize individual vocabulary, voice, and speech properties,
or:
2. The user tailors their own voice pronunciation and vocabulary to the program's expectations.
Starting with the oldest and simplest single purpose programs have limited and preset vocabularies. Also they usually have the lowest prices:

1. The simplest single purpose command and control utility programs with a limited preset command word vocabulary may require you to train your own voice pronunciation to the program's expectations. As they have preset, and/or limited, command word vocabularies.
A few will allow you to add a limited number of your own voice command controls. The simplest may add your commands phonetically and require no further voice training.
2. The simplest single purpose discrete speech dictation [Note Pads] programs with a limited preset vocabulary may require you to train your voice pronunciation to the program's expectations. As they have preset, and or limited, voice dictation word vocabularies.
A few will allow you to add a limited number of your own voice vocabulary words. The simplest may add your vocabulary words phonetically and require no further voice training. Others will require the addition of your own voice pronunciation to train the program.
Not All Dictation Programs Are Created Equal.
All dictation programs will enter spoken text into your computer. Not all enter spoken text to the same place or application in your computer. Some will only enter to Clipboard, Note pad, Write, Word, or WordPerfect. The most versatile will enter text directly into any Windows compatible application.
Not all dictation programs will provide Speech Command and Control over your computer applications and operations. This an important option that can substantially reduce or eliminate the necessity of keyboard and mouse input. Full voice input [without the necessity of keyboard or mouse] is currently available to: * IBM ViaVoice Gold , and Dragon Systems Dragon Dictate *.
In speech dictation you talk text into your computer. To read text out of your computer you need a text -to- speech [screen reader] synthesizer. Compatibility between programs is critical as these are two different multi-tasking programs, with two different purposes. Together they form a feedback loop with the user.

Use Your Browser To Return

GOOD HUNTING AND ENJOY!
SUPERADAPTOID.
* Top * | * ACSP Home * | * SuperAdaptoid Column *

01-01-98