• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2007.11a
• /
• pp.439-442
• /
• 2007

강인한 얼굴 인식 시스템을 만들기 위해서는 안경의 제거가 중요한 요소이다. 이를 위해서는 뛰어난 성능의 안경 검출 방법이 필수적이다. 본 논문에서는 안경의 유무 판단에 관한 새로운 방법을 제안한다. 영상은 조명 부분과 반사부분의 곱으로 이루어져 있다. 얼굴의 경우 안경 고유의 반사계수와 얼굴 고유의 반사계수가 다른 점에 착안하여 anisotropic smoothing 방법을 이용하여 입력 얼굴 영상에서의 반사 부분을 얻고, 이를 이용하여 안경의 반사 부분을 얼굴의 반사부분에서 검출한 뒤 이진화한다. 이후, 이진화 된 안경 픽셀 수를 이용하여 안경의 유무를 판단한다.

• Annual Conference on Human and Language Technology
• /
• 2023.10a
• /
• pp.437-442
• /
• 2023

최근에 등장한 대규모 언어 모델은 다양한 언어 처리 작업에서 놀라운 성능을 발휘하고 있다. 그러나 이러한 모델의 크기와 복잡성 때문에 모델 경량화의 필요성이 대두되고 있다. Pruning은 이러한 경량화 전략 중 하나로, 모델의 가중치나 연결의 일부를 제거하여 크기를 줄이면서도 동시에 성능을 최적화하는 방법을 제시한다. 본 논문에서는 한국어 언어 모델인 Polyglot-Ko에 Wanda[1] 기법을 적용하여 Pruning 작업을 수행하였다. 그리고 이를 통해 가중치가 제거된 모델의 Perplexity, Zero-shot 성능, 그리고 Fine-tuning 후의 성능을 분석하였다. 실험 결과, Wanda-50%, 4:8 Sparsity 패턴, 2:4 Sparsity 패턴의 순서로 높은 성능을 나타냈으며, 특히 일부 조건에서는 기존의 Dense 모델보다 더 뛰어난 성능을 보였다. 이러한 결과는 오늘날 대규모 언어 모델 중심의 연구에서 Pruning 기법의 효과와 그 중요성을 재확인하는 계기가 되었다.

• Polymer(Korea)
• /
• v.29 no.2
• /
• pp.127-134
• /
• 2005

The relationships among mechanical properties, rheological properties, and morphology by reactive extrusion based on commingled pckaging film wastes contains polypropylene (PP) pckaging film system [PP/polyethylene (PE)/aluminum (Al)/poly(ethylene terephthalate) (PET)] and Nylon packaging film system[Nylon/PE/linear-low density polyethylene (LLDPE)] were investigated to improve the compatibility and toughness of these wastes using various compatibilizers such as ethylene vinylacetate (EVA), styrene-ethylene/butylene-styrene triblock copolymer (SEBS), styrene-ethylene/butylene-styrene-graft-maleic anhydride copolymer (SEBS-g-MA), polyethylene-graft-maleic anhydride (PE-g-MA), polypropylene-graft-maleic anhydride (PP-g-MA) , polyethylene-graft-acrylic acid (PE-g-AA) and polypropylene-graft-acrylic acid (PP-g-AA). Compared with simple melt blend system, the blends showed improvement of about $50\%$ increase in physical properties when SEBS and EVA were added. However, SEBS-g-MA thermoplastic elastomer which is highly reactive with amine terminal group of nylon, resulted in about $200\%$ increase in impact strength. This compatibilization effect resulted from the increase of interfacial adhesion and the reduction of domain size of dispersed phase in PP/Nylon blend system.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.42 no.1
• /
• pp.59-65
• /
• 2005

This paper proposes a novel TSK fuzzy modeling algorithm. Various approaches to fuzzy modeling when noise or outliers exist in the data have been presented but they are approaches to degrade effects of outliers or large noise by using loss function in the cost function mainly. The proposed algorithm is the modified version of noise clustering algorithm, and it adopts the method that does not use loss function, but method to cluster noise in a class. Noise clustering is a prototype-based clustering algorithm and it has no capability to regress. It conducts clustering of data first, and then conducts fuzzy regression. There are many algorithms to obtain parameters of premise and consequent part simultaneously, but they need to adapt the parameters obtained for more accurate approximation. In this paper, fuzzy regression is conducted with clustering by modifying noise clustering algorithm. We propose the algorithm that parameters of the premise part and the consequent part are obtained simultaneously, and the parameters obtained are not needed to adapt. We verify the proposed algorithm through simple examples and evaluate the test results compared with existing algorithms. The proposed algorithm shows robust performance against noise and it is easy to implement.

• The Journal of the Acoustical Society of Korea
• /
• v.20 no.3
• /
• pp.56-62
• /
• 2001

In this paper, we adopt a missing data theory to speech recognition. It can be used in order to maintain high performance of speech recognizer when the missing data occurs. In general, hidden Markov model (HMM) is used as a stochastic classifier for speech recognition task. Acoustic events are represented by continuous probability density function in continuous density HMM(CDHMM). The missing data theory has an advantage that can be easily applicable to this CDHMM. A marginalization method is used for processing missing data because it has small complexity and is easy to apply to automatic speech recognition (ASR). Also, a spectral subtraction is used for detecting missing data. If the difference between the energy of speech and that of background noise is below given threshold value, we determine that missing has occurred. We propose a new method that examines the reliability of detected missing data using voicing probability. The voicing probability is used to find voiced frames. It is used to process the missing data in voiced region that has more redundant information than consonants. The experimental results showed that our method improves performance than baseline system that uses spectral subtraction method only. In 452 words isolated word recognition experiment, the proposed method using the voicing probability reduced the average word error rate by 12％ in a typical noise situation.

• The Journal of the Acoustical Society of Korea
• /
• v.38 no.6
• /
• pp.670-677
• /
• 2019

As the presence of background noise in acoustic signal degrades the performance of speech or acoustic event recognition, it is still challenging to extract noise-robust acoustic features from noisy signal. In this paper, we propose a combined structure of Wasserstein Generative Adversarial Network (WGAN) and MultiTask AutoEncoder (MTAE) as deep learning architecture that integrates the strength of MTAE and WGAN respectively such that it estimates not only noise but also speech features from noisy acoustic source. The proposed MTAE-WGAN structure is used to estimate speech signal and the residual noise by employing a gradient penalty and a weight initialization method for Leaky Rectified Linear Unit (LReLU) and Parametric ReLU (PReLU). The proposed MTAE-WGAN structure with the adopted gradient penalty loss function enhances the speech features and subsequently achieve substantial Phoneme Error Rate (PER) improvements over the stand-alone Deep Denoising Autoencoder (DDAE), MTAE, Redundant Convolutional Encoder-Decoder (R-CED) and Recurrent MTAE (RMTAE) models for robust speech recognition.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.22 no.6
• /
• pp.681-686
• /
• 2012

This paper studied the speech parameters less affected by the human emotion for the development of the robust emotional speech recognition system. For this purpose, the effect of emotion on the speech recognition system and robust speech parameters of speech recognition system were studied using speech database containing various emotions. In this study, mel-cepstral coefficient, delta-cepstral coefficient, RASTA mel-cepstral coefficient, root-cepstral coefficient, PLP coefficient and frequency warped mel-cepstral coefficient in the vocal tract length normalization method were used as feature parameters. And CMS (Cepstral Mean Subtraction) and SBR(Signal Bias Removal) method were used as a signal bias removal technique. Experimental results showed that the HMM based speaker independent word recognizer using frequency warped RASTA mel-cepstral coefficient in the vocal tract length normalized method, its derivatives and CMS as a signal bias removal showed the best performance.

• The Journal of the Acoustical Society of Korea
• /
• v.25 no.4
• /
• pp.151-158
• /
• 2006

In this paper, we proposed a new audio fingerprint extraction method, based on Philips' music retrieval algorithm, which uses the energy difference of neighboring filter-bank and probabilistic characteristics of music. Since Philips method uses too many filter-banks in limited frequency band, it may cause audio fingerprints to be highly sensitive to additive noises and to have too high correlation between neighboring bands. The proposed method improves robustness to noises by reducing the number of filter-banks while it maintains the discriminative power by representing the energy difference of bands with 2 bits where the quantization levels are determined by probabilistic characteristics. The correlation which exists among 4 different levels in 2 bits is not only utilized in similarity measurement. but also in efficient reduction of searching area. Experiments show that the proposed method is not only more robust to various environmental noises (street, department, car, office, and restaurant), but also takes less time for database search than Philips in the case where music is highly degraded.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.40 no.4
• /
• pp.700-708
• /
• 2015

A software birthmark is the set of characteristics of a program which can be used to identify the program. Many researchers have studied on detecting theft of java programs using some birthmarks. In case of Android apps, code obfuscation techniques are used to protect the apps against reverse-engineering and tampering. However, attackers can also use the obfuscation techniques in order to conceal a stolen program. A birthmark (feature) of an app can be alterable by code obfuscations. Therefore, it is necessary to detect Android app theft based on the birthmark which is resilient to code obfuscation. In this paper, we propose an effective Android app birthmark and app theft detection through the proposed birthmark. By analyzing some obfuscation tools, we have first selected parameter and the return types of methods as an adequate birthmark. Then, we have measured similarity of target apps using the birthmarks extracted from the apps, where some target apps are not obfuscated and the others obfuscated. The measurement results show that our proposed birthmark is effective for detecting Android app theft even though the apps are obfuscated.

• The Journal of the Acoustical Society of Korea
• /
• v.40 no.2
• /
• pp.130-138
• /
• 2021

This paper presents an effective structure by applying various normalization to Convolutional Neural Networks (CNN) for seismic event classification. Normalization techniques can not only improve the learning speed of neural networks, but also show robustness to noise. In this paper, we analyze the effect of input data normalization and hidden layer normalization on the deep learning model for seismic event classification. In addition an effective model is derived through various experiments according to the structure of the applied hidden layer. As a result of various experiments, the model that applied input data normalization and weight normalization to the first hidden layer showed the most stable performance improvement.