• Annual Conference on Human and Language Technology
• /
• 2020.10a
• /
• pp.37-41
• /
• 2020

본 논문에서는 Joint CTC/Attention 모델에 CTC ratio scheduling을 이용한 end-to-end 한국어 음성인식을 연구하였다. Joint CTC/Attention은 CTC와 attention의 장점을 결합한 모델로서 attention, CTC 단일 모델보다 좋은 성능을 보여주지만, 학습이 진행될수록 CTC가 attention의 학습을 저해하는 요인이 된다. 본 논문에서는 이러한 문제를 해결하기 위해, 학습 진행에 따라 CTC의 비율(ratio)를 줄여나가는 CTC ratio scheduling 방법을 제안한다. CTC ratio scheduling를 이용하여 학습한 결과물은 기존 Joint CTC/Attention, 단일 attention 모델 대비 좋은 성능을 보여주는 것을 확인하였다.

• The Journal of the Acoustical Society of Korea
• /
• v.21 no.5
• /
• pp.471-477
• /
• 2002

Recently, lip-reading has been studied actively as an auxiliary method of automatic speech recognition(ASR) in noisy environments. However, almost of research results were obtained based on the database constructed in indoor condition. So, we dont know how developed lip-reading algorithms are robust to dynamic variation of image. Currently we have developed a lip-reading system based on image-transform based algorithm. This system recognize 22 words and this word recognizer achieves word recognition of up to 53.54%. In this paper we present how stable the lip-reading system is in environmental variance and what the main variant factors are about dropping off in word-recognition performance. For studying lip-reading robustness we consider spatial valiance (translation, rotation, scaling) and illumination variance. Two kinds of test data are used. One Is the simulated lip image database and the other is real dynamic database captured in car environment. As a result of our experiment, we show that the spatial variance is one of degradations factors of lip reading performance. But the most important factor of degradation is not the spatial variance. The illumination variances make severe reduction of recognition rates as much as 70%. In conclusion, robust lip reading algorithms against illumination variances should be developed for using lip reading as a complementary method of ASR.

• The Journal of the Acoustical Society of Korea
• /
• v.20 no.1
• /
• pp.50-55
• /
• 2001

This paper presents a korean isolated words recognition algorithm which used new endpoint detection method, auditory model, 2D-DCT and new distance measure. Advantages of the proposed algorithm are simple hardware construction and fast recognition time than conventional algorithms. For comparison with conventional algorithm, we used DTW method. At result, we got similar recognition rate for speaker dependent korean isolated words and better it for speaker independent korean isolated words. And recognition time of proposed algorithm was 200 times faster than DTW algorithm. Proposed algorithm had a good result in noise environments too.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.14 no.1
• /
• pp.77-92
• /
• 2020

When automatic speech recognition (ASR) is provided as a cloud service, it is easy to collect voice and application domain data from users. Harnessing these data will facilitate the provision of more personalized services. In this paper, we demonstrate our transfer learning-based knowledge service that built with the user-generated data collected through our novel system that deliveries personalized ASR service. First, we discuss the motivation, challenges, and prospects of building up such a knowledge-based service-oriented system. Second, we present a Quadruple Transfer Learning (QTL) method that can learn a classification model from a source domain and transfer it to a target domain. Third, we provide an overview architecture of our novel system that collects voice data from mobile users, labels the data via crowdsourcing, utilises these collected user-generated data to train different machine learning models, and delivers the personalised real-time cloud services. Finally, we use the E-Book data collected from our system to train classification models and apply them in the smart TV domain, and the experimental results show that our QTL method is effective in two classification tasks, which confirms that the knowledge transfer provides a value-added service for the upper-layer mobile applications in different domains.

• Journal of the Ergonomics Society of Korea
• /
• v.17 no.3
• /
• pp.23-39
• /
• 1998

Several recent aircraft accidents occurred due to goal conflicts between human and machine actors. To facilitate the management of the cockpit activities considering these observations. a computational aid. the Agenda Manager (AM) has been developed for use in simulated cockpit environments. It is important to know pilot intentions performing cockpit operations accurately to improve AM performance. Without accurate knowledge of pilot goals or intentions, the information from AM may lead to the wrong direction to the pilot who is using the information. To provide a reliable flight simulation environment regarding goal conflicts. a pilot goal communication method (GCM) was developed to facilitate accurate recognition of pilot goals. Embedded within AM, the GCM was used to recognize pilot goals and to declare them to the AM. Two approaches to the recognition of pilots goals were considered: (1) The use of an Automatic Speech Recognition (ASR) system to recognize overtly or explicitly declared pilot goals. and (2) inference of covertly or implicitly declared pilot goals via the use of an intent inferencing mechanism. The integrated mode of these two methods could overcome the covert goal mis-understanding by use of overt GCM. And also could it overcome workload concern with overt mode by the use of covert GCM. Through simulated flight environment experimentation with real pilot subjects, the proposed GCM has demonstrated its capability to recognize pilot intentions with a certain degree of accuracy and to handle incorrectly declared goals. and was validated in terms of subjective workload and pilot flight control performance. The GCM communicating pilot goals were implemented within the AM to provide a rich environment for the study of human-machine interactions in the supervisory control of complex dynamic systems.

• The Journal of the Acoustical Society of Korea
• /
• v.20 no.4
• /
• pp.35-41
• /
• 2001

This paper describes Linear Discriminant Analysis and common vector extraction for speech recognition. Voice signal contains psychological and physiological properties of the speaker as well as dialect differences, acoustical environment effects, and phase differences. For these reasons, the same word spelled out by different speakers can be very different heard. This property of speech signal make it very difficult to extract common properties in the same speech class (word or phoneme). Linear algebra method like BT (Karhunen-Loeve Transformation) is generally used for common properties extraction In the speech signals, but common vector extraction which is suggested by M. Bilginer et at. is used in this paper. The method of M. Bilginer et al. extracts the optimized common vector from the speech signals used for training. And it has 100％ recognition accuracy in the trained data which is used for common vector extraction. In spite of these characteristics, the method has some drawback-we cannot use numbers of speech signal for training and the discriminant information among common vectors is not defined. This paper suggests advanced method which can reduce error rate by maximizing the discriminant information among common vectors. And novel method to normalize the size of common vector also added. The result shows improved performance of algorithm and better recognition accuracy of 2％ than conventional method.

• Journal of Korea Multimedia Society
• /
• v.11 no.11
• /
• pp.1471-1479
• /
• 2008

In this paper, we propose a technique to improve Out-Of-Vocabulary(OOV) rejection algorithms in variable vocabulary recognition system which is much used in ASR(Automatic Speech Recognition). The rejection system can be classified into two categories by their implementation method, keyword spotting method and utterance verification method. The utterance verification method uses the likelihood ratio of each phoneme Viterbi score relative to anti-phoneme score for deciding OOV. In this paper, we add speaker verification system before utterance verification and calculate an speaker verification probability. The obtained speaker verification probability is applied for determining the proposed variable-confidence threshold. Using the proposed method, we achieve the significant performance improvement; CA(Correctly Accepted for keyword) 94.23%, CR(Correctly Rejected for out-of-vocabulary) 95.11% in office environment, and CA 91.14%, CR 92.74% in noisy environment.