Sunday, March 5, 2006

Is Vemma Good For Inflammation

TTS: Text To Speech

AT&T Labs Research

AT&T AT & T opens the doors of his laboratory research . TTS technology is the creation, from a typed text, speech audible.
Introduced in 1998, the second generation of TTS has been proven. Yep, this is definitely not science fiction. This technology is increasingly used and is found in many applications. To take just the best known: the cinema! Imagine that the film "Red Planet" (Warner Bros. - 09/11/2000) benefited from AT & T TTS: the female voice of the spaceship that transports our hero on Mars, it also incorporated the male voice uniforms of astronauts. Obviously there are certainly
most common uses, such as voice announcing the subway stations, bus or train in our public transit ... but I am appealing to your scholarship to enlighten me on this point.
There are also interpretation software called Screen Reader "used for the visually impaired that render the oral text on a webpage.

Well, it's all well and good, but how does it work?

In three steps, from the moment you click 'SPEAK'

1) Preprocessing the
The Role of the pre-processor is to divide the text typed in words, symbols and phrases strongly linked, and remove stray characters (spaces, line-end characters, special characters ...). Using regular expressions, the preprocessor detects what we call wordins tokens ("string of characters between two spaces"), but also specific expressions such as URLs, dates, phone numbers or punctuation. When an expression is detected, the pre-processor launches analyzer particular, in order to identify the expression in the different units present.

2) Morphological analysis
In a first step, the analysis will try to identify the various syntactic categories attributed to each language. In a second step, these are the words that will be given a list of possible lexical types (name, verb, ...).

3) Parsing
This module has the task of deciding what syntactic category corresponds to each linguistic unit, given the context of the sentence. To do this, the analyzer, random type, combining two models. The first model is that of language and the second model is the word that is introduced into the calculation.

The synthesis software must then make an acoustic prosody: the pronunciation of a sentence means to determine the melody (prosody). Two information must be generated: the fundamental frequency and duration of each phoneme.
And finally use an algorithm type MBROLA (Multi-Band Re-synthesis OverLap Add) that works in two phases:
First, it extracts a database acoustic units corresponding to the list of phonemes to be synthesized, and applies to them, one by one, prosody previously calculated. In a second step, these units are concatenated, and a spectral smoothing is undertaken at their borders, so will remove discontinuities. The result is a voice synthesis of very good quality, which still remains distinguishable from the natural voice.
multitel-eLite (These explanations are derived from a PDF available on the site MULTITEL TTS Group , you can also download a competitor to that developed by AT & T, but it must be installed on its position).

I invite you to visit the online demo Tool AT & T ! Restitution of words is incredible fidelity both in terms of pronunciation as the accent. Remains the key of a complete sentence which is not equivalent to a human, but it's not far off (the punctuation is well respected in all cases).

Here are some tips to tweak the program a bit with lines of code, and the complete list of codes usable.

Enjoy, it's free and fun!

0 comments:

Post a Comment