• 한국컴퓨터정보학회논문지
• /
• 제23권2호
• /
• pp.99-107
• /
• 2018

This study started from the public interest purpose to help accessibility for the information acquisition of the vulnerable groups due to visual difficulties such as the elderly and the visually impaired. In this study, the server resources are minimized and implemented in most of the user smart phones. In addition, we implement a method to gather necessary information by collecting only pattern information by utilizing crawl & clipping without having to visit the site of the information of the various sites having the data necessary for the user, and to have it in the server. Especially, we applied the TTS(Text-To-Speech) service composed of smart phone apps and tried to develop a unified customized information collection service based on voice-based information collection method.

• 한국음향학회지
• /
• 제22권8호
• /
• pp.727-739
• /
• 2003

본 논문에서는 HM-Net (Hidden Markov Network)을 다양한 태스크에의 적용과 화자의 특성을 효과적으로 나타내기 위해 HM-Net 음성인식 시스템에 MLLR (Maximum Likelihood Linear Regression) 적응방법을 도입하였으며, HM-Net 학습 알고리즘을 개량하여 회귀클래스 생성방법을 제안한다. 제안방법은 PDT-SSS (Phonetic Decision Tree-based Successive State Splitting)알고리즘의 문맥방향 상태분할에 의한 상태레벨 공유를 이용한 방법이다. 즉, 문맥방향의 각 상태에 적응화자 음성데이터에 포함된 문맥정보를 분할하여 적응화될 음소환경을 결정하는 것이다. 따라서 제안방법은 새로운 화자로부터 문맥정보와 적응화 데이터의 발성 양에 의존하여 결정된 많은 적응 파라미터들을 (평균, 분산) 자유롭게 제어할 수 있게 된다. 제안방법의 유효성을 확인하기 위해 국어공학센터 (KLE) 452 데이터와 항공편 예약관련 (YNU200) 연속음성을 대상으로 인식실험을 수행한 결과, 음소인식, 단어인식, 연속음성인식에 대해서, 평균 34∼37%, 평균 9%, 평균 20%의 성능 향상을 각각 보였다. 또한 적응화 데이터의 양에 따른 인식성능 비교에서 제안방법을 적용한 인식 시스템이 적응 데이터의 양이 적은 경우에도 향상된 인식률을 보여 MLLR 적응방법의 특성을 만족하였다. 따라서 MLLR 적응방법을 도입한 HM-Net 음성인식 시스템에 제안한 회귀클래스 생성방법이 유효함을 확인할 수 있었다.

• Journal of the Korean Association of Oral and Maxillofacial Surgeons
• /
• 제35권4호
• /
• pp.221-228
• /
• 2009

The purpose of the study was to investigate the characteristics of the pronunciation of Korean vowels in patients with class III malocclusion. 11 adult male patients with class III malocclusion(mean ages 22.3 years) and four adult males with normal occlusion(mean ages 26.5 years) were selected for the analysis of eight Korean monophthongs /ㅣ, ㅔ, ㅐ, ㅏ, ㅓ, ㅗ, ㅡ, ㅜ/. The values and relationships of F1, F2 and F3 were derived from the stable section of target vowel in each sentence, and the analysis using formant plots and vowel triangles' distance and area was conducted to find the features of two groups' vowel distributions. Consequently, it was identified that the pronunciation of males patients with class III malocclusion showed high values of F1 in the low vowels, high values of F2 in the back vowels, and remarkably low position of /ㅏ/. The vowel triangle suggested that the triangle areas of male patients with class III malocclusion were shown wider vertically and narrower horizontally than those of males with normal occlusion. These characteristics could reflect the structural features of class III malocclusion such as the prognathic mandible, low tongue position, and advancement of back position of the tongue.

• Journal of Audiology & Otology
• /
• 제25권1호
• /
• pp.14-21
• /
• 2021

Background and Objectives: There is growing interest in the use of the Level-specific (LS) CE-Chirp^® stimulus in auditory brainstem response (ABR) due to its ability to produce prominent ABR waves with robust amplitudes. There are no known studies that investigate the test-retest reliability of the ABR to the LS CE-Chirp^® stimulus. The present study aims to investigate the test-retest reliability of the ABR to the LS CE-Chirp^® stimulus and compare its reliability with the ABR to standard click stimulus at multiple intensity levels in normal-hearing adults. Subjects and Methods: Eleven normal-hearing adults participated. The ABR test was repeated twice in the same clinical session and conducted again in another session. The ABR was acquired using both the click and LS CE-Chirp^® stimuli at 4 presentation levels (80, 60, 40, and 20 dBnHL). Only the right ear was tested using the ipsilateral electrode montage. The reliability of the ABR findings (amplitudes and latencies) to the click and LS CE-Chirp^® stimuli within the same clinical session and between the two clinical sessions was calculated using an intra-class correlation coefficient analysis (ICC). Results: The results showed a significant correlation of the ABR findings (amplitude and latencies) to both stimuli within the same session and between the clinical sessions. The ICC values ranged from moderate to excellent. Conclusions: The ABR results from both the LS CE-Chirp^® and click stimuli were consistent and reliable over the two clinical sessions suggesting that both stimuli can be used for neurological diagnoses with the same reliability.

• 대한청각학회지
• /
• 제25권1호
• /
• pp.14-21
• /
• 2021

Background and Objectives: There is growing interest in the use of the Level-specific (LS) CE-Chirp^® stimulus in auditory brainstem response (ABR) due to its ability to produce prominent ABR waves with robust amplitudes. There are no known studies that investigate the test-retest reliability of the ABR to the LS CE-Chirp^® stimulus. The present study aims to investigate the test-retest reliability of the ABR to the LS CE-Chirp^® stimulus and compare its reliability with the ABR to standard click stimulus at multiple intensity levels in normal-hearing adults. Subjects and Methods: Eleven normal-hearing adults participated. The ABR test was repeated twice in the same clinical session and conducted again in another session. The ABR was acquired using both the click and LS CE-Chirp^® stimuli at 4 presentation levels (80, 60, 40, and 20 dBnHL). Only the right ear was tested using the ipsilateral electrode montage. The reliability of the ABR findings (amplitudes and latencies) to the click and LS CE-Chirp^® stimuli within the same clinical session and between the two clinical sessions was calculated using an intra-class correlation coefficient analysis (ICC). Results: The results showed a significant correlation of the ABR findings (amplitude and latencies) to both stimuli within the same session and between the clinical sessions. The ICC values ranged from moderate to excellent. Conclusions: The ABR results from both the LS CE-Chirp^® and click stimuli were consistent and reliable over the two clinical sessions suggesting that both stimuli can be used for neurological diagnoses with the same reliability.

• 전자공학회논문지B
• /
• 제33B권3호
• /
• pp.169-175
• /
• 1996

In this paper, a new parallel neural network system that performs dynamic competitive learning is proposed. Conventional learning mehtods utilize the full dimension of the original input patterns. However, a particular attribute or dimension of the input patterns does not necessarily contribute to classification. The proposed system consists of parallel neural networks with the reduced input dimension in order to take advantage of the information in each dimension of the input patterns. Consensus schemes were developed to decide the netowrks performs a competitive learning that dynamically generates output neurons as learning proceeds. Each output neuron has it sown class threshold in the proposed dynamic competitive learning. Because the class threshold in the proposed dynamic learning phase, the proposed neural netowrk adapts properly to the input patterns distribution. Experimental results with remote sensing and speech data indicate the improved performance of the proposed method compared to the conventional learning methods.

• 대한치과교정학회지
• /
• 제5권1호
• /
• pp.69-73
• /
• 1975

This is case report of true class III malocclusion with macroglossia is corrected by glossectomy in 13 years female patient. After orthodontic treatment, the patient is bound to glossectomy because the corrected condition is relapsed to the previous condition due to relatively enlarged tongue compared with the original dental arch. By the interpretation of the cephalogram and model analysis, it is approved that the growth pattern and direction are normal range and mandible is located anterioly to the cranium. The results are follows: 1. We could treat the true Cl III malocclusion. 2. We could prevent the relapse of the treated condition by the surgical intervention, such as partial glossectomy. 3. Sensory, speech, swallowing and so other functions after the operation have been with in normal limit without any serious complications or seguellae.

• 전자공학회논문지B
• /
• 제31B권12호
• /
• pp.108-115
• /
• 1994

Most pattern classifiers have been designed based on the ML (Maximum Likelihood) training algorithm which is simple and relatively powerful. The ML training is an efficient algorithm to individually estimate the model parameters of each class under the assumption that all class models in a classifier are statistically independent. That assumption, however, is not valid in many real situations, which degrades the performance of the classifier. In this paper, we propose a minimum-error-rate training algorithm based on the MAP (Maximum a Posteriori) approach. The algorithm regards the normalized outputs of the classifier as estimates of the a posteriori probability, and tries to maximize those estimates. According to Bayes decision theory, the proposed algorithm satisfies the condition of minimum-error-rate classificatin. We apply this algorithm to NPM (Neural Prediction Model) for speech recognition, and derive new disrminative training algorithms. Experimental results on ten Korean digits recognition have shown the reduction of 37.5% of the number of recognition errors.

• 공학교육연구
• /
• 제21권5호
• /
• pp.3-9
• /
• 2018

This article examined the effects of self-reflecting journal writing in speaking class on academic performance of science and engineering students. To assess the effect, 27 science and engineering students from the "Speech and Life" class were asked to keep a self-reflecting journal. Pre and post-intervention surveys were conducted, followed by the analysis of learning effect and satisfaction. In addition to the pre and post-intervention surveys, an additional survey on speaking ability was conducted at the same time and the change of the students' ability was assessed. Results showed that after writing self-reflection journals, participants' learning effect and satisfaction has increased, and their speaking performance was also improved.

• 한국통신학회논문지
• /
• 제31권7C호
• /
• pp.691-699
• /
• 2006

VoIP(Voice over Internet Pratocol)와 같은 IP 네트워크망에서는 패킷 지연, 지터, 패킷 손실 등의 이유로 QoS(Quality of Service)를 보장받지 못하기 때문에, 패킷 손실을 은닉하는 방법에 대한 연구는 필수적이다. IP망에서 사용되는 대부분의 저전송률 음성부호화기는 자체적으로 패킷 손실 은닉(PLC: Packet Loss Concealment) 알고리즘을 사용하고 있지만, 예측 기법에 기반한 양자화 특성상 패킷 손실 이후에도 에러가 전파되는 문제가 있다. 또한, 손실된 패킷의 음성신호 특성을 고려하지 않고 과거 파라미터값을 반복시키는 기존 PLC 방법은 그 구현은 쉽지만 천이구간에서의 합성신호의 음질이 심각히 저하된다. 본 논문에서는 패킷 손실 환경에서 랩신호 특성에 따른 에러전파 영향을 정량적으로 분석하고 그 결과를 토대로 보간법 기반의 새로운 PLC 알고리즘을 제안한다. 제안한 알고리즘은 파라미터별로 음성신호의 특성을 고려해 선택적으로 보간법을 적용하고, 예측 필터의 메모리를 효과적으로 갱신한다. 성능평가 결과, 제안한 알고리즘은 VoIP에서 널리 사용되는 G.729 의 기존 PLC 알고리즘에 비해 다양한 FER 환경에서 성능이 향상되었다.