Get best of the week

History of Speech Synthesizers: The Computer Age

Earlier we talked about how speech synthesizers appeared - we talked about the first mechanical and electrical installations. We continue the discussion with what happened in the era of mainframes.


Photo Kate Ter Haar / CC BY

What else is on our blog on Habré -> Reading for the weekend: 10 materials about musical easter eggs, hidden tracks, rhythm games and free audio samples

Hybrid computer systems


It is believed that the first computer system for speech synthesis appeared in Japan. It was developed by a team of specialists from the Electrotechnical Laboratory (ETL), which the Ministry of Communications founded in the 19th century. In 1956, its employees introduced the ETL Mark III transistor computer - one of the first computing systems with built-in memory. It was difficult to find a more suitable place for developing a speech synthesizer in the country.

Engineers began their work in 1963with building a polymer model of the human vocal tract. For two years, experts studied the physical processes occurring in it - they paid special attention to the role of the nose in the coloring of speech. After collecting the necessary data, they turned to colleagues from Hitachi, who helped design an analog computer based on the ladder logic . It included 71 operational amplifiers and 22 multipliers, which were resistor networks with photoconductors and neon lamps. The system connected to the IBM 7090 with Fortran control software.

. 1967 . — , . — .


The head of the acoustic department of the ETL laboratory, Eiichi Matsui, took the ideas of his colleagues in the late 60s and designed a fully digital speech synthesizer for the IBM S / 360 computer. To model the characteristics of the vocal tract, he used the Fourier transform. Each second of synthesized speech required 20 seconds of pre-processing. But despite this shortcoming, the project attracted the attention of the international scientific community.

Research in this area was started by experts from Nagoya University, also located in Japan. One of the scientists of the university proposed a new method of speech synthesis - linear prediction . It takes into account several time-varying parameters: the period of the fundamental tone, the average volume and the factors associated withthe transfer function of the vocal tract.

Such synthesizers calculated the necessary parameters from digitized speech, recorded them in memory, and then reproduced sound waves using the DAC. The system did not require large computing power, so the linear prediction method quickly found application in many speech recognition systems.

. Votrax « » . 1974 (Donald Sherman), ( ), .

— , «». , . . , , .


In 1976, Raymond Kurzweil, an American inventor, introduced his “ reading machine ” to the world . A box the size of an outdoor office printer scanned documents, recognized them, and read them out loud. In the same year, a “talking” calculator, LPC Speech +, was released . It was built on the basis of the first commercial microcircuit with an integrated voice synthesizer.

Two years later, Texas Instruments released Speak & Spell , a children's educational toy with a built-in speech synthesizer. In the memory of the device 200 words were put, which children often pronounce incorrectly. The child typed a word on the keyboard, and the toy reproducedhis. The product became an absolute hit - it was even used by musicians in their projects. The voice of Speak & Spell can be heard on Kraftwerk's Computer World album.

Speak & Spell, albeit in an updated form, is on the shelves today. In a way, the gadget was the first smart digital device. He is even called the "forefather" of personal computers. But we will talk about how speech synthesis systems developed on a PC next time.


« Hi-Fi»:

:
:
« »:
Sporth — live-
—

More articles:


All Articles