|
Text-to-Speech SynthesisIntroductionA Text-To-Speech (TTS) synthesizer is a computer-based system that should be able to read any text aloud, whether it was directly introduced in the computer by an operator or scanned and submitted to an Optical Character Recognition (OCR) system. Let us try to be clear. There is a fundamental difference between the system we are about to discuss here and any other talking machine (as a cassette-player for example) in the sense that we are interested in the automatic production of new sentences. This definition still needs some refinements. Systems that simply concatenate isolated words or parts of sentences, denoted as Voice Response Systems, are only applicable when a limited vocabulary is required (typically a few one hundreds of words), and when the sentences to be pronounced respect a very restricted structure, as is the case for the announcement of arrivals in train stations for instance. In the context of TTS synthesis, it is impossible (and luckily useless) to record and store all the words of the language. It is thus more suitable to define Text-To-Speech as the automatic production of speech, through a grapheme-to-phoneme transcription of the sentences to utter. At first sight, this task does not look too hard to perform. After all, is not the human being potentially able to correctly pronounce an unknown sentence, even from his childhood ? We all have, mainly unconsciously, a deep knowledge of the reading rules of our mother tongue. They were transmitted to us, in a simplified form, at primary school, and we improved them year after year. However, it would be a bold claim indeed to say that it is only a short step before the computer is likely to equal the human being in that respect. Despite the present state of our knowledge and techniques and the progress recently accomplished in the fields of Signal Processing and Artificial Intelligence, we would have to express some reservations. As a matter of fact, the reading process draws from the furthest depths, often unthought of, of the human intelligence. 1. Automatic Reading : what for ?Each and every synthesizer is the result of a particular and original imitation of the human reading capability, submitted to technological and imaginative constraints that are characteristic of the time of its creation. The concept of high quality TTS synthesis appeared in the mid eighties, as a result of important developments in speech synthesis and natural language processing techniques, mostly due to the emergence of new technologies (Digital Signal and Logical Inference Processors). It is now a must for the speech products family expansion. Potential applications of High Quality TTS Systems are indeed numerous. Here are some examples : Telecommunications services. TTS systems make it possible to access textual information over the telephone. Knowing that about 70 % of the telephone calls actually require very little interactivity, such a prospect is worth being considered. Texts might range from simple messages, such as local cultural events not to miss (cinemas, theatres,... ), to huge databases which can hardly be read and stored as digitized speech. Queries to such information retrieval systems could be put through the user's voice (with the help of a speech recognizer), or through the telephone keyboard (with DTMF systems). One could even imagine that our (artificially) intelligent machines could speed up the query when needed, by providing lists of keywords, or even summaries. In this connection, AT&T has recently organized a series of consumer tests for some promising telephone services [Levinson et al. 93]. They include : Who's Calling (get the spoken name of your caller before being connected and hang up to avoid the call), Integrated Messaging (have your electronic mail or facsimiles being automatically read over the telephone), Telephone Relay Service (have a telephone conversation with speech or hearing impaired persons thanks to ad hoc text-to-voice and voice-to-text conversion), and Automated Caller Name and Address (a computerized version of the "reverse directory"). These applications have proved acceptable, and even popular, provided the intelligibility of the synthetic utterances was high enough. Naturalness was not a major issue in most cases. Language education. High Quality TTS synthesis can be coupled with a Computer Aided Learning system, and provide a helpful tool to learn a new language. To our knowledge, this has not been done yet, given the relatively poor quality available with commercial systems, as opposed to the critical requirements of such tasks. Aid to handicapped persons. Voice handicaps originate in mental or motor/sensation disorders. Machines can be an invaluable support in the latter case : with the help of an especially designed keyboard and a fast sentence assembling program, synthetic speech can be produced in a few seconds to remedy these impediments. Astro-physician Stephen Hawking gives all his lectures in this way. The aforementioned Telephone Relay Service is another example. Blind people also widely benefit from TTS systems, when coupled with Optical Recognition Systems (OCR), which give them access to written information. The market for speech synthesis for blind users of personal computers will soon be invaded by mass-market synthesisers bundled with sound cards. DECtalk (TM) is already available with the latest SoundBlaster (TM) cards now, although not yet in a form useful for blind people. Talking books and toys. The toy market has already been touched by speech synthesis. Many speaking toys have appeared, under the impulse of the innovative 'Magic Spell' from Texas Instruments. The poor quality available inevitably restrains the educational ambition of such products. High Quality synthesis at affordable prices might well change this. Vocal Monitoring. In some cases, oral information is more efficient than written messages. The appeal is stronger, while the attention may still focus on other visual sources of information. Hence the idea of incorporating speech synthesizers in measurement or control systems. Multimedia, man-machine communication. In the long run, the development of high quality TTS systems is a necessary step (as is the enhancement of speech recognizers) towards more complete means of communication between men and computers. Multimedia is a first but promising move in this direction. Fundamental and applied research. TTS synthesizers possess a very peculiar feature which makes them wonderful laboratory tools for linguists : they are completely under control, so that repeated experiences provide identical results (as is hardly the case with human beings). Consequently, they allow to investigate the efficiency of intonative and rhythmic models. A particular type of TTS systems, which are based on a description of the vocal tract through its resonant frequencies (its formants) and denoted as formant synthesizers, has also been extensively used by phoneticians to study speech in terms of acoustical rules. In this manner, for instance, articulatory constraints have been enlightened and formally described. 2. How does a machine read ?From now on, it should be clear that a reading machine would hardly adopt a processing scheme as the one naturally taken up by humans, whether it was for language analysis or for speech production itself. Vocal sounds are inherently governed by the partial differential equations of fluid mechanics, applied in a dynamic case since our lung pressure, glottis tension, and vocal and nasal tracts configuration evolve with time. These are controlled by our cortex, which takes advantage of the power of its parallel structure to extract the essence of the text read : its meaning. Even though, in the current state of the engineering art, building a Text-To-Speech synthesizer on such intricate models is almost scientifically conceivable (intensive research on articulatory synthesis, neural networks, and semantic analysis give evidence of it), it would result anyway in a machine with a very high degree of (possibly avoidable) complexity, which is not always compatible with economical criteria. After all, flies do not flap their wings ! Figure 1 introduces the functional diagram of a very general TTS synthesizer. As for human reading, it comprises a Natural Language Processing module (NLP), capable of producing a phonetic transcription of the text read, together with the desired intonation and rhythm (often termed as prosody), and a Digital Signal Processing module (DSP), which transforms the symbolic information it receives into speech. But the formalisms and algorithms applied often manage, thanks to a judicious use of mathematical and linguistic knowledge of developers, to short-circuit certain processing steps. This is occasionally achieved at the expense of some restrictions on the text to pronounce, or results in some reduction of the "emotional dynamics" of the synthetic voice (at least in comparison with human performances), but it generally allows to solve the problem in real time with limited memory requirements. Programmer: Mike Perry
|
