News register
Search section
Search area
News Type
News type
Search date
Search word OR
List of related articles
Contents of related articles
No info was found
List of related articles
Contents of related articles
No info was found
View details
Information

06/10/2019 HYU News > Academics > 이달의연구자

Title

[Excellent R&D] Innovation of Robotic Technology for Conversational Voice Recognition

Professor Chang Joon-hyuk (Department of Electronic Engineering)

김민재

Copy URL / Share SNS

http://www.hanyang.ac.kr/surl/tiT2

Contents
With the emergence of the Fourth Industrial Revolution, countries and large conglomerates are concentrating on developing technologies combined with artificial intelligence (AI). Hanyang University is also putting significant emphasis on collaboration with major companies and fostering students to become experts knowledgeable about the current trend. In fact, Hanyang University does not stop at doing research and writing theses; rather, it comes up with actual products based on the theses and provides the technology to companies for further progressive development. Professor Chang Joon-hyuk (Department of Electronic Engineering), one of the professors who practices the university’s pursuit, has successfully invented robotic technology for conversational voice recognition which can be applied to our daily life.
 
Professor Chang Joon-hyuk (Department of Electronic Engineering) has invented robotic technology that recognizes people’s conversational voices.

Chang believes that voice is the most natural medium in human communication as it involves emotional delivery and indicates overall mood. Doing research where robots or computers recognize such types of voice is an extremely interesting field of study. Compared to the past when the performance of voice recognition was poor due to lack of technological development, nowadays there have been greater results in a short period of time thanks to the use of AI. As this voice recognition algorithm was applied to smartphones, automobiles, and robots, it has had a considerable impact on the market and industry. “I am happy that our research resulted in higher performance which led to the enhancement of the algorithm, but I feel a sense of pride that this research has strengthened the competence of Korean industries and created more potential occupations,” said Chang.
 
Chang is currently working on creating a 100 percent AI-based voice recognition device. In particular, when the most natural interface tool is voice, he wanted to equip an ordering and replying mechanism in the robot that enables human-robot conversation. Chang played a pivotal role in creating an AI robot named Cloy for Incheon International Airport. This robot works as an information desk that answers customers’ questions. However, it has earned a low performance level as the environmental noise is too loud; thus, it is still going through enhancement procedures. In addition, Chang is working on assignments appointed by Hyundai Automobile. Since typing while driving a car is dangerous, Hyundai wants to apply the voice recognition system there as well. Currently, Hyundai is borrowing the engines from Kakao corporation and Amazon, but it wants to equip its own engine. For the capability reinforcement of Hyundai, Chang participated in the joint development project based on an engine made from Hanyang University.

At the same time, he is carrying out a project that improves automobile After Service (AS). It is important to distinguish specific disorders within the car, and this particular maintenance duty can be only done by a proficient mechanic. However, the device conceived of by one of Chang’s student who is in the Doctorate Program discerns the possible problems in the automobile down to three issues simply by listening to the noise generated from the car so that the mechanic can easily proceed with his maintenance work. According to Chang, "Since such a technique is very efficient in terms of time and economic aspects, it has received high evaluations from Hyundai, my student was offered special employment and a monetary reward, which I think is a win-win situation for both the company and the school.”
 
Chang is currently working on various projects with renowned large conglomerates that have already led to a win-win situation for both companies and the university.
When inventing an AI algorithm, three major factors are to be considered. First, it is important to collect and handle data to establish a fine data base. Second is developing a deep learning AI algorithm that is almost as same as human thoughts or that even surpass human functionality, and then properly combining voice/sound perception with the data. Last but not least, apply the result and make an actual product. With the basis of these factors, Hyundai Automobile collected data by intentionally breaking all parts of automobiles to construct a database.

Nonetheless, since deep learning is available only through big data, the database provided by Hyundai was extremely limited. Therefore, Chang decided to apply the method of data amplification which involves mathematical analysis and signal processing. The mathematical approach utilizes the concept of probability that first composes a probability model and then deducts abundant data following that model. On the other hand, the signal processing approach collects data from a specific environment. It then applies the signal characteristics of a variety of different environments to the given result from the specific environment. This way, data can be limitlessly reproduced.

Chang suggested, “Thanks to such an algorithm, the performance rate increased from 50 percent to approximately 90 percent. This is revolutionary from the perspective of timeliness and economy.” Chang’s lab has also invented an AI speaker called PLUTO that has been exhibited in one of the most renowned home appliance exhibitions. This speaker perceives sound well and can understand a voice from a long distance and in a noisy environment. Samsung, LG, and Hyundai Automobiles are in joint development of this huge project which is expected to be extremely prosperous.
 
Speech recognition: strong voice recognition is necessary in actual environments where various noises and reverberation exist; therefore, applying diverse machine learning algorithms improves the performance level in actual environments.
(Photo courtesy of ASML Lab website)

Chang wants students to be more aware of the fast-changing generation in terms of technological developments. “Simply listening to lectures and doing assignments is not enough. Students are encouraged to realize the current industry trends and developing their personal capabilities in accordance with the advancing society. Also, students’ ultimate goal should not be limited to employment in major companies. They should continuously put effort into enhancing their competence in accordance with the fourth industrial revolution,” concluded Chang.



Kim Min-jae       fhffl5781@hanyang.ac.kr
Photos by Kim Joo-eun
Copy URL / Share SNS

0 Comments