RSS
Логотип
Баннер в шапке 1
Баннер в шапке 2

Speech Technology Center Group Center for Speech Technologies

Company

width=200px
Speech Technology Center Group (part of the Sberbank ecosystem) is a global developer of intelligent speech technologies, face recognition, a technology expert in the field of artificial intelligence and machine learning. One of the few companies in the world that creates and develops both biometric modalities: face and voice. Voice fake detection technologies and speech recognition from the CST group occupy leading positions in the world rankings NIST, ASVspoof Challenge, VOiCES, CHiME Challenge. Solutions are CST in demand in 70 countries of the world.

Owners:
Gazprombank (GPB)
Digital Horizon
Revenue millions Ths. rub

Number of employees

Owners

+ Rosneft of the Tax Code
+ Nowfinteh
+ Oleg Mikhailovich Vaksman

The company's clients include commercial organizations, various power structures, law enforcement agencies, government agencies, whose activities are especially important for the high-quality transmission, registration and processing of speech information.

The company has licenses for the development and production of special and military equipment, including using information representing state secrets.

The quality level of the company's work is confirmed by the certificate of the international quality standard of ISO-9001:2008 and the Russian QMS GOST RV 15.002-2003 and GOST R ISO 9001-2001.

  • Speech recognition - the conversion of Russian oral speech into electronic text.
  • High-quality synthesis of Russian speech.
  • Voice biometrics: verification and identification of identity by phonograms of oral speech.
  • Development of modern voice self-service systems for IVR contact centers
  • Speech Recognition - Recognize voice commands in any language.
  • Noise cleaning and increased speech intelligibility.
  • Forensic identification, diagnosis of the appearance and nationality of the person by phonograms of oral speech.

Serial products

The key areas of work of the CST are the development of human biometrics systems, identification and verification (for example, for security systems), voice accounting systems (the Mexican police use the development of CST in the database of criminal voices), speech synthesis and analysis systems (voice self-service), recording and signal analysis (). call centers

The CST produces more than 20 products:

  • The IKAR Lab phonogram research complexes are in service with almost all forensic centers and laboratories of law enforcement agencies of the Russian Federation, countries of near and far abroad, and the Interstate Aviation Committee.
  • The Tral-M phonogram automation and rapid research system is increasingly recognized by experts and security specialists.
  • Digital voice recorders of the Gnome series and multi-channel Forget-Me-Not systems have become standards of high-quality sound recording, which, along with rich functionality, allows them to be successfully used in a wide variety of fields of activity.
  • STC Call-Logger is the first Russian development to monitor the operation of call center operators.
  • Multichannel digital tape recorders P-424M and P-425M adopted and replaced their analog predecessors on ships of the Russian Navy.
  • Together with NPO "Device" (SPb), a multi-channel non-kinematic digital tape recorder module for aircraft was developed.
  • The scope of Tishina embedded boards designed for noise cleaning and increased speech intelligibility in communication channels is constantly expanding.
  • Nestor's speech documentation systems provide prompt processing of materials of meetings and meetings in the higher bodies state power of Russia.
  • Autonomous Cinderella-series noise cleaning devices and Sound Cleaner software complexes showed high efficiency when working with low-quality, noisy and distorted signals
  • The voice notification system "Rupor" allows you to bring the necessary information to the employees of the enterprise, subscribers, personnel of the departments of the Ministry of Internal Affairs, the Ministry of Emergencies and the Civil Defense as soon as possible.

CST innovation

In July 2011, CST innovation became a resident of Skolkovo. This is a 100% subsidiary of the St. Petersburg "Center for Speech Technologies," created specifically for the new project of "multimodal biometric systems." According to CNews, the executive director of "MDG-Innovation" Innokenty Dementiev, in addition to the already received exemption from income tax, benefits for the payment of insurance premiums and a simplified export regime, the company is counting on a grant of "several million dollars "CST[1].

The development of the "MDG-Innovation" will be an identification technology that allows you to identify a person by voice or face image. Potential customers in the company are called the spheres of state, corporate security and telecommunications.

The company considers the starting development unique in scale and functionality: "it will be able to store multimillion-dollar databases of biometric data (samples of voices, photo images), find target personalities in any communication channels and video files, analyze media of all types in real time and build instant response systems."

The software is directly aimed at the security forces. "Identification of the criminal, suspect or any other target person will be carried out on the basis of an analysis of audio recordings (for example, telephone conversations), photographs or video material obtained at the facility or crime scene," the CST says. - It is also possible to connect additional identification modules: by fingerprint, iris, etc. '.

'We are cooperating with the Interior Ministry and the FSB on our other products, 'says CNews Dementiev. "Previously, the departments became interested in the new development, we received letters from the Ministry of Internal Affairs about this, but its implementation is not a matter of the very near future, after the completion of the work, a number of examinations will still be required."

"As far as I know, the CST is really successfully cooperating with the security forces, it is more difficult with projects in the consumer market," said Vyacheslav Borilin, vice president of the developer of video conference systems Spirit. - The fact is that commercial customers often do not have specially trained employees with the system, as in the case of special services, but untrained and unmotivated personnel. "

Recall that the large-scale voice biometric system implemented by the "Center for Speech Technologies" in Mexico has been in operation since the beginning of 2010. 'Future development will allow identifying a person in real time by two biometric modalities,' explains Dementiev. - Currently, Mexico has implemented a system of phonocounts, that is, it collects, stores and searches only by the vote base. In the framework of our new project, firstly, a system will be developed that will allow identifying a person not by one biometric characteristic, but by two, namely, by voice and by face. Such a system will make it most likely to identify a person's identity. In addition, all processes in the system will proceed automatically, which will allow even specialists without expert qualifications to use our system.

In addition to law enforcement agencies, strategic objects, venues for mass events, etc. are called a niche for their new system in the CST. Dementiev calls its main competitors NEC, Morpho (a subsidiary of the Safran group, which deals with aviation and security systems).

Performance indicators

2021: Revenue growth by 53.2% to RUB 3,993 million

At the end of 2021, the revenue of the Speech Technology Center Ltd. of Companies amounted to 3,993 million rubles, an increase of 5.32% compared to 2020, which allowed it to take 115th place in the TAdviser100: The largest IT companies in Russia 2022.

History

2022

Prize of the Government of the Russian Federation for the development of voice input for radiologists

Scientists of the Center for Diagnostics and Telemedicine of the DZM and the CST group of companies received the Prize of the Government of the Russian Federation for the adaptation and introduction of speech recognition technology for the preparation of protocols for X-ray and ultrasound studies. This was announced on December 26, 2022 by Zdrav.Expert representatives of the Center for Diagnostics and Telemedicine of DZM. Read more here.

Sberbank ceased to be a shareholder of the CST

As TAdviser found out, Sberbank's subsidiary Digital Assets LLC, which was the founder and co-founder of a number of companies related to the bank's digital business, withdrew from these assets in May. In particular, from the companies "Center for Speech Technologies," "Cloud Technologies" (operates on the market under the SberCloud brand), "Okko," "Evotor." In addition, Digital Assets LLC left the Sound company. Read more here.

2020

Revenue - 2,607 million rubles

At the end of 2020, the revenue of the Speech Technology Center Ltd. company amounted to 2,607 million rubles.

Strategic Partnership With VARA Technology To Promote Group Products In India

VARA Technology (VARA) will offer specialized facial and voice biometrics solutions from the CST group of companies to corporate customers in India. The strategic partnership provides for the expansion of the VARA portfolio with products and solutions of the CST group, which will be presented to financial and public institutions, transport enterprises, telecommunications service providers. This was announced on September 28, 2020 in the CST group of companies. Read more here.

Speech technologies CST at the heart of Naumen's AI solution for classifying calls

On June 22, 2020, the company NAUMEN introduced a solution for robotic classifying incoming calls, which completely replaces - IVR menu contact center in and is installed on top of any communication platform. The product based on AI the platform Naumen Erudite speech technologies and the CST group of companies will help contact centers improve customer experience and reduce costs by reducing the classification time of each call to 20 seconds. More. here

Recognition of speech recognition technology CST the best in the world

The technology of diarization and speech recognition, created by the CST group of companies (part of the ecosystem) Sberbank, was recognized as the best at the international CHiME Speech Separation and Recognition Challenge (CHiME-6[2] This was TAdviser reported on May 7, 2020 in the CST.

The technology was recognized for recognizing English speech from several microphones in a natural environment. The CST group showed the best test results in the most difficult task[3] competition[4]significantly outperforming competitors.

CHiME organizers offer teams from all over the world various tasks that become more difficult with each competition. In CHiME-5, the contestants solved the so-called cocktail party problem - recognizing the spontaneous speech of several announcers in conditions of partial overlap of speech and noise, that is, in a typical communication situation at a party. This block required work with segmented (already allocated) speech. The peculiarity of the CHiME-6 was that the contestants were asked to solve a similar problem, but working with unsegmented speech, while - with speech overlap up to 20%. It is on solving this, the most difficult task that the CST team has focused.

Entries for the competition were made at 20 dinners in actual homes at parties where people cooked, ate, washed dishes, socialised freely and emotionally, joked and laughed. For recognition, simultaneous speech of 2-4 people, reverb and intense noise are difficult here - the ringing of devices, water pouring from the tap, the hum of the air conditioner, steps, laughter. The goal of the participants is to create a recognition system that "listens" to the recordings and produces a complete decryption with the least number of errors. The CST team took first place:

On the graph: the results of the competition, in the columns - the number of errors made. Photo: chimechallenge.github.io

To do this, an algorithm was developed for allocating speech segments for each of the announcers, and a complex of several neural networks of different architectures was created, distinguishing between different announcers, implementing bimforming (the effect of targeting microphones to a particular announcer) and directly recognizing speech.

In addition to the CST group, scientific teams from all over the world participated in the competition: well-known IT companies - Toshiba and a number of others, and major leading universities in the field of speech technologies: Johns Hopkins University (USA), University of Science and Technology of China, Technical University Brno (Czech Republic), etc.

File:Aquote1.png
The CST group has been creating, developing and improving speech technologies for 30 years. In 2020, the CHiME-6 had the most difficult task - working with unsegmented speech. High-quality speech recognition of different announcers, while interrupted by noise, allows you to withdraw services from the category of innovative in everyday use, improving business and simplifying our lives. Thus, high-quality processing of unsegmented speech will allow, for example, to conduct competent logging of meetings, where several speakers speak at once, and intelligent speech analytics will automate the work: contact centers recognize spontaneous speech, classify voice calls, identify script compliance, draw conclusions about client satisfaction and the quality of dialogue, which means significantly optimize the work of modern contact centers, and retail. e-commerce telecom The recognition of the CST group in this international competition is not just our personal victory, but a landmark event for the entire industry, and we are pleased to bring the solution to the speech recognition problems that the strongest teams from all over the world are working on to a different level, adequately representing their key competencies in the global market, - comments the CEO of the CST group of companies. Dmitry Dyrmovsky
File:Aquote2.png

{{quote 'The task of CHiME is to ensure the exchange of experience of the strongest teams from all over the world and to move forward the solution of global problems in the field of speech recognition. And we welcome the achievements of the CST group of companies in this area, "said John Barker, a representative of the University of Sheffield (UK), a member of the CHiME Challenge organizing committee. }}

2019

Closing the transaction of Sberbank's acquisition of 51% of the CST shares

Sberbank, Gazprombank and Digital Horizon closed a deal in which Sberbank acquired a 51% stake in the Speech Technology Center. This was reported on August 5, 2019 by Sberbank.

The shareholders formed a board of directors, which included four representatives of Sberbank, two representatives of Gazprombank and one representative of Digital Horizon. Konstantin Kruglov, head of SberDevices, was elected chairman.

The Center for Speech Technologies, which has become part of the Sberbank ecosystem, specializes in the development of artificial intelligence technologies, including speech recognition and synthesis and computer vision. Thanks to the deal, the company will have access to the resources and competencies of the two largest financial organizations in Russia, as well as to the international expertise of Digital Horizon, which will allow CST developments to compete in the global technological market.

File:Aquote1.png
Sberbank has already begun to use the development of CST in its products and technologies. For example, including on their basis, a digital TV presenter Elena was created, presented by us in the spring of 2019. I am confident that working with CST specialists will help complement and improve the products and services of our growing ecosystem. An important task of the team will be to enter the international market, where the company has every opportunity to take a significant market share in the segment of speech technologies, competing with international players, - said Konstantin Kruglov, chairman of the board of directors of the CST and head of SberDevices.
File:Aquote2.png

File:Aquote1.png
Dmitry Dyrmovsky, CEO of the CST Group'With the closure of the CST deal, has acquired a strategic partner, synergy with which sets a new powerful impetus for our development. CST products based on face recognition work in sports, transport, create an infrastructural city throughout the country. Expertise and support of one of the leading banks in the world will serve as a driver for the implementation of our goals in the Russian and global markets.
File:Aquote2.png

"Center for Speech Technologies" came under the control of Sberbank

Logo CST

On April 12, 2019, Sberbank informed TAdviser about the conclusion of a deal with Gazprombank and venture capital company Digital Horizon, in which it acquired from Gazprombank a 51% stake in the CST group of companies, a Russian developer of biometric technologies. Digital Horizon also entered the capital of the CST. The financial terms of the transaction were not disclosed. It is planned to close it by the end of May 2019. At the same time, Gazprombank remains a strategic shareholder of the CST and will continue to actively participate in the further development of the company.

According to Sberbank, shareholders will update the composition of the company's board of directors, which will include 4 representatives from Sberbank, including the chairman of the board of directors, as well as two representatives from Gazprombank and one from Digital Horizon.

The deal will allow the CST to gain access to the resources and competencies of the two largest financial organizations in Russia, including in the field of artificial intelligence and big data. In addition, the involvement of the expertise of the international team Digital Horizon will provide the CST with additional competitive advantages in the global market, the volume of which is projected at $40 billion already in 2022, Sberbank emphasized.

File:Aquote1.png
For Sberbank, the deal represents a natural step towards digital transformation and building a biometric platform for the bank's growing ecosystem. In the future, various services will be created on the basis of this platform, which will allow you to translate the format of interaction with customers to a qualitatively new level of comfort and information security. In turn, Sberbank's expertise and experience in working with AI and big data will allow the CST to increase the use of voice technologies in the country, as well as replicate Russian technologies at the international level and claim world leadership, "commented Stanislav Kuznetsov, Deputy Chairman of the Board of Sberbank.
File:Aquote2.png

File:Aquote1.png
The CST acquires another serious partner, synergy with which will provide a powerful impetus for the development of technologies and products of our company. We gain access to the expertise and competencies of one of the largest retail banks in the world. The CST is one of the main players in the biometric market, but in such a team we expect to become world leaders in the future for several years, "said Dmitry Dyrmovsky, CEO of the CST group.
File:Aquote2.png

File:Aquote1.png
Gazprombank Group has successful experience in investing in promising high-tech companies, which confirms our long-term joint work with the CST. We intend to continue to maintain a strategic focus on the development of this area within the framework of the group's work in order to introduce advanced biometric developments into the daily work of the bank in the interests of our numerous retail and corporate clients, - said Dmitry Sauers, Deputy Chairman of the Management Board of Gazprombank.
File:Aquote2.png

File:Aquote1.png
This is a historic deal not only for the development of biometrics, but also for the venture capital investment industry in Russia, "says Alan Waxman, co-founder of Digital Horizon. - Thanks to the long-term vision of Gazprombank, which saw the potential of the CST over 10 years ago, Sberbank, as the flagship of the Russian digital economy, is building a biometric platform unique by world standards. As an investor in the CST, we will continue our work to position the company on the international market and further increase its shareholder value.
File:Aquote2.png

2016

Revenue growth of 39.8%

In 2016, the CST revenue amounted to 1.219 billion rubles, which is 39.8% higher than the same indicator in 2015. In the ranking "TAdviser100: The largest IT companies in Russia" at the end of 2016, the CST took 60th place.

Partnership Agreement with Egyptian Falcon Group

In late 2016, CST entered into a partnership with Falcon, a security solutions provider based in Egypt. The official agreement was signed by Falcon President, Sharif Khaled and Commercial Director of CST, Akop Mkhitaryan in the presence of Egypt's Ambassador to France, Ehab Badawi.

Falcon and the CST plan to develop joint solutions to control access and monitor infrastructure facilities such as airports, stadiums, cultural heritage sites, hotels and conference centers. The main markets for such systems will be the Middle East and Africa. The CST views this region as a priority, and has already introduced several products with support for the Egyptian dialect of Arabic. In this regard, the CST partnership with Falcon represents a unique opportunity to combine the technological advances and expertise of the two companies.

"Due to the need to take measures to prevent terrorist acts and crimes, the demand for high-tech security equipment is steadily growing. This issue is relevant both for border control points and strategically important facilities, and for places of mass stay of people: stadiums, universities and tourist zones. And for countries looking to increase the flow of tourists, innovative equipment plays a crucial role. In light of these trends, biometric technologies are extremely promising, as they provide an unprecedented level of security, "comments the commercial director of the CST, Akop Mkhitaryan
.

QMS CST complies with ISO 9001:2015

The quality management system of the CST was certified for compliance with the international standard ISO 9001:2015. A special transitional audit was conducted by Det Norske Veritas (DNV).

DNV auditors analyzed and evaluated all the company's business processes, while no inconsistencies were identified and several cases of good practice were noted, these are: management's promotion of risk-oriented thinking; involvement of employees in improving QMS efficiency; development and implementation of a unified financial model of the company based on the investment plan, sales funnel, project budgets; accumulation and dissemination of knowledge in various areas of activity (processes, products, technologies) in the Confluence knowledge base.

Certification of the quality management system according to ISO 9001-2015 means that the CST have advanced methods for managing product development and production, service, and customers can be sure that the company's products and services are safe, reliable and of high quality.

2015

25 years of CST

CST presented projects and products

On October 23, 2015, it became known about the presentation of CST 's solutions at the Moscow Expocenter as part of an exhibition on import substitution in contact centers[5]

The company's specialists demonstrated the possibilities of using CST solutions in credit institutions:

  • Voice Self-Service System VoiceNavigator
  • Platform for voice and photo identification of clients in remote service VoiceKey
  • Audio and video recording and analysis systems, including voice and on-screen analytics tools for contact centers
  • Automatic information systems by digital and analog communication channels with the function of speech synthesis and recognition of the Horn series

Expocenter (2015)

A number of the company's projects implemented for banks and financial institutions have become innovative:

  • TransCreditBank has introduced an automated service for finding branches and ATMs,
  • NPF "Trust" protected customers through the use of VoiceGrid X,
  • UBRD has implemented voice analytics systems in the information and reference center.

2011: Sale to Gazprombank

In August 2011, it became known that Gazprombank acquired a stake in the Speech Technology Center (CST), which belonged to the Quadriga Capital Russia investment fund (part of the Quadriga Capital international fund) [6]In addition, the bank acquired part of the shares of the founders of the CST. The parties did not disclose the financial terms of the transaction.

The CST revenue in 2010 amounted to about $18 million, the EBITDA margin has not changed over the past few years and remains at the level of 25%. Earlier, analysts estimated the business of the CST according to the multiplier of 10-12 EBITDA, thus, the cost of the entire company today can be $45-54 million. According to approximate estimates of Finam, the cost of the CST may be about $50 million. The amount of the transaction with Gazprombank, according to a source familiar with the negotiations, amounted to $32 million.

CEO of the CST Mikhail Hitrov says that the investor needed the company to reach "new frontiers," and to solve "large-scale problems."

"We are the global technology leader in knowledge-intensive markets such as speech synthesis and recognition, voice biometrics, speech analytics, etc.," says Hitrov. - Moreover, the competitive advantages of the company are its own scientific school (one of the largest in the world) and the level of development. Our task is to maximize this innovative potential in the world and Russian markets. "

According to Vladimir Rumyantsev, head of the department of telecommunications, media and high technologies of Gazprombank, "voice biometrics, synthesis and speech recognition are one of the most promising areas in the high-tech sector of the economy."

"We consider it very important and necessary for the bank to help reach a new level of scale of operations of the Russian company, which is already one of the world leaders in the development of new products in this highly competitive segment," says Rumyantsev.

As a result of the transaction, the company will retain its current management. "The new shareholders supported the course that the CST are developing," the source adds. Whether the funds received by the founders as a result of the transaction will be directed to the development of the company, the source does not specify.

2009: Involvement in politicians' wiretapping scandal in Colombia

In 2009, a scandal erupted in Colombia: the local intelligence service - Departamento Administrativo de Seguridad (DAS) - was accused of intercepting telephone conversations between opposition politicians, journalists, and even Supreme Court judges. It was alleged that the DAS acted on the orders of the country's president, Alvaro Urib. The scandal was dubbed the "Colombian Watergate" (by analogy with the Watergate scandal that erupted in the United States in the 1970s, when President Richard Nixon was caught wiretapping his political opponents).

In connection with the start of the hype, the head of DAS Felipe Munoz held a press conference in the Colombian capital Bogota with Sergei Koval, the chief expert of the CST company, presented as an independent expert. Koval reported that he studied about 20 samples of audio recordings related to the aforementioned leaks. The Russian expert came to the conclusion that they were made with fundamentally different equipment than DAS uses, respectively, the Colombian special service has nothing to do with these leaks.

Munoz was pleased with this conclusion, noting that there is an illegal market for wiretapping equipment that is beyond control. However, it later turned out that Koval's conclusions were incorrect. Former senior DAS employee William Remero admitted that he was engaged in wiretapping on the direct instructions of Maria del Pilar Hurtado, who headed the DAS until 2008. Ultimately, the auditions were presented to President Uribe.

Prosecutors have brought serious charges against the DAS leadership. Maria Pilar Hurtado fled the country and was granted asylum in Panama. She later had to return to Colombia, where she was sentenced to 14 years. DAS itself was disbanded in 2011[7].

2003: Quadriga Capital Fund acquires 35% in CST

In 2003, Quadriga Capital became a shareholder in the CST, buying 35%. Since then, he has repeatedly unsuccessfully tried to sell this share. The founders of the company - 4 people, including Mikhail Hitrov - also intended to find a buyer for their packages for several years. The deal was discussed, for example, with Nokia and Nuance, the world leader in speech recognition.

A CST source says the founders did not want the company to dissolve into an international giant, becoming a competence center for voice technology. That is why Gazprombank was chosen as an investor.

1993: Immigrants from the Institute of Long-Distance Communications form a company

The Center for Speech Technologies (CST, SpeechPro trademark) was founded in 1993 by a number of former employees of the Leningrad Institute of Long-Distance Communications, including Sergei Koval and Mikhail Hitrov.

In Soviet times, the Institute of Long-Distance Communications (Dalsvyaz) was the main coordinator of the work carried out in the interests of the State Security Committee (KGB) to intercept telephone conversations. Koval was responsible for the work of the acoustic laboratory dealing with the voice recognition system. Developments in this area began in the late 1940s in the "sharashka" in Marfino near Moscow.

The Academy of Sciences of the USSR was also connected to solving the problems of recognizing voices. To this end, in the 1980s, the KGB allowed the Academy of Sciences to purchase the first batch of personal computers in the USSR. But with the collapse of the USSR, work in this area stopped, and Dalsvyaz employees had to look for other classes[8].

But the CST team managed to continue doing what they loved on a commercial basis. The company received profitable contracts from the FSB and by 2000 CST of people worked in 350 - almost the same as in the best years it was in Dalsvyaz.

The CST even managed to create a national system for recognizing voices, conceived in Soviet times, only this happened in the USSR, but in Mexico. The system, launched in 2008, allows the Mexican authorities to search for people using samples of biometric information: samples of voice, photos, etc. Law enforcement officers, prisoners and other categories of citizens who have to face the authorities (for example, this must be done when obtaining a driver's license) are required to take voice samples for the system.

Membership in organizations

The CST is:

  • co-founder of the Russian Speech Technologies consortium;
  • member of the Russian Biometric Society;
  • member of the Council of the innovative educational program of St. Petersburg State University (national project "Education");
  • co-editor of the International Speech Biometric Standard from the Russian Federation;

member of the Public Organization and the Regional Association of Employers "Union of Industrialists and Entrepreneurs of St. Petersburg" (NGO SPP SPb and ROR SPP SPb).

Links

g-l "Expert": Someone's voice sang to me (11.2009)

Notes