The present article aims to reveal historical background of a concept «neural networks» and its importance in information technology. The author carried out literature review according to the development of the concept «neural network». As the product of Artificial Intelligence, speech recognition process has been described in the given article.
Keywords: computational theory, neural networks, speech processing, speech recognition, artificial intelligence.
The humanity is passing to the fourth industrial revolution and witnessing the booming of artificial intelligence, robotics, quantum computing, and other advances of sophisticated technology. We all have used Siri as a personal voice assistant when Apple company launched this app. This is a question-answer system, which has been adapted for the iOS operating system. Other speech recognition technologies such as Amazon’s Alexa, Cortana, and Google Assistant are changing the way people interact with their devices, homes, cars, and jobs. All of these advances of informational technologies makes our life easier and fascinating. These apps answer to our questions and follow our commands. They refer to ANI (artificial narrow intelligence). We must remember that narrow intelligence does not mention low intelligence.
Speech recognition relates to the computer science, computer engineering and computational linguistics. Words of the spoken language can be identified by applying speech recognition algorithm.
English theoretical physicist, cosmologist Stephen Hawking once said: «Success in creating AI would be the biggest event in human history. Unfortunately, it might also be the last, unless we learn how to avoid the risks». Having mentioned about the advantages of using AI scientist also noted its dangers, like powerful autonomous weapons, or new ways for the few to oppress the many [1]. There will be also certain risks such as lack of transparency and loss of control.
At the beginning, it is better to look through the historical development of the concept «neural network».
It turns out that 150 years have passed since the emergence of the concept of» neural network» in our society. When the concept «neural networks» first appeared it only applied to the study of neurons in the brain and spinal cord. Since that time this concept enlarged it’s usage up to information technology that humanity use in wide range across our global village. Then there in science introduced artificial neural networks. Research in the field of artificial neural networks went through different stages.
Researches started to speak about artificial neural networks since 1940. The first scholars who researched this new field were American researchers Warren S.McCulloch and Walter H.Pitt. In 1943, they wrote an article «A Logical Calculus of the ideas Immanent in Nervous Activity» [2]. This work can be considered as the starting point of neural network investigation. These researchers showed that any logical and arithmetic algorithms could be implemented using neural networks. They confirmed that the usage of logic and computation can contribute to understand neural, and thus mental, activity.
The next who showed interest in this field was Canadian psychologist Donald O.Hebb. His found out how the function of neurons contributed to psychological processes such as learning process. He is best known for his theory of Hebbian learning, which he introduced in his classic 1949 work «The Organization of Behavior«[3]. In science, he is well known as the father of neuropsychology and neural networks. Before McCulloch and Pitts nobody had used the mathematical notion of computation as an ingredient in a theory of mind and brain.
Frank Rosenblatt introduced the second stage of the neural network in 1958. He called it as single-layer perceptron. In this model, neurons were connected by variable weights: the output is one of the weighted sums of its input is above threshold and zero if it is below. The first implementation of the perceptron was not by software, because it was prolonged back then. Rosenblatt decided to build his device were weights were implemented by variable resistors, and learning the weights was done by electric motors that turned the knobs on the resistor [4].
A group of researchers as Tommi Kinnunen from the University of Carnegie Mellon (USA), the University of Illinois (USA), the University of Oregon (USA), the University of Eastern Finland, etc., that have achieved significant success in speech recognition practice in the world. Foreign universities have achieved a lot in comparison with the CIS countries.
Researchers from Kazakhstan also paid an interest in this new field of neural networks. Since the active use of computer technologies in the 21-century Kazakhstani scientists have defended a lot articles are written, a number of thesis. PhD doctor О. J. Mamyrbaev who is working at the Institute оf Infоrmatiоn and Cоmputatiоnal Technоlоgies in our country has contributed to the development of artificial intelligence studying speech recognition. He is the author of many interesting articles like «Automatic gender іdentіfіcatіоn іn speech recоgnіtіоn by genetic algorithm» [5]. «Automatic recognition of Kazakh speech using Deep Neural Networks», «Systematic review and analysis of voice identification features» [6], «Usage of MFCC algorithm in the process of speech recognition» For example, scholar proposed identication of autоmatіc gender speech according to the gender usіng speech sіgnals that can develоp speech encоdіng, analysіs and synthesisіs. From these lists, it can be seen that he has written a lot of articles devoted to speech recognition with the help of neural networks.
Professors of Eurasian National University after L. N. Gumilyov A. Sharipbay and scientist from Al-Farabi Kazakh National University U. A. Tukeyev also engaged in studying neural networks in our country.
Nowadays problem of speech recognition has become the object of information technologies. As the speech is the most common and natural phenomenon of human communication it is vital to use computer system as it is considered to be the effective input of information and management of mobile systems. The progress in this area significantly speeds up the process of communication.
Wouter Gevaert, Georgi Tsenov, Valeri Mladenov depicted the speech recognition process in the following way:
Fig. 1. Speech recognition process
The first block, represents acoustic environment and transduction equipment like microphone, preamplifier and AD-converter. The second block is can deal with acoustic problems. The third block must be capable of extracting speech specific features of the pre-processed signal. This can be done with techniques like cepstrum analysis and the spectrogram. The fourth block tries to classify the extracted features, relates the input sound to the best fitting sound in a known ‘vocabulary set’, and represents this as an output [7].
Speech recognition algorithm is now widely applied in all business spheres like recording conference calls and physical meetings, language translation for travelers, dictate medical reports. The public sector has adopted A. I. technologies for a variety of purposes. Status of AI industry is increasing in the world.
In conclusion, we would like to say that notwithstanding the problems mankind is facing nowadays developing the speech recognition process, it’s future will be promising opt to speech is a very subjective phenomenon.
References:
- Hawking S. Creating AI Could Be the Biggest Event in the History of Our Civilization. [Electronic resource] https://futurism.com/hawking-creating-ai-could-be-the-biggest-event-in-the-history-of-our-civilization
- McCulloch, W.S., Pitts W. H. A Logical Calculus of the Ideas Immanentin Nervous Activity. Bulletin of Mathematical Biophysics 7, 115–133. Reprinted inMcCulloch 1964. P. 16–39.
- Shaw G. L. Donald Hebb: The Organization of Behavior. In: Palm G., Aertsen A. (eds) Brain Theory. Springer, Berlin, Heidelberg. — 1986.
- Domingos P. The Master Algorithm: How the Quest for the Ultimate Learning Machine Will Remake Our World. Basic Books, Inc., USA. –2018.
- Mamyrbaev О. J. Elektr énergetïkaliq komplekstiñ turaqtiliq jumisiniñ aqparattiq jüyesin qurw. Nauchnaya diskussiya: innovatsii v sovremennom mire. sb. st. po materialam XLVIII mezhdunar. nauch.-prakt. konf. — № 4 (47). Chast' II. — M., Izd. «Internauka», 2016. — 220 s.
- Mamyrbaev О. J. Autоmatіc gender іdentіfіcatіоn іn speech recоgnіtіоn by genetіc algоrіthm. Vestnik Almatinskogo universiteta Energetiki i Svyazi. Spetsial'nyy vypusk 2018. — P.120–129.
- Gevaert W., Tsenov G., Mladenov V. Neural Networks used for Speech Recognition. Journal of automatic control, University of Belgrade, Vol. 20:1–7, 2010.
