분절 특징 HMM을 이용한 영어 음소 인식

윤영선,Yun, Young-Sun 한국정보과학회 2002 정보과학회논문지 : 소프트웨어 및 응용 Vol.29 No.3

본 논문에서는 여러 프레임 특징으로 표현되는 분절 특징(segmental feature) 표현 방법을 제안하고, HMM 개념 위에서 음향학적 모델과 그 알고리즘을 개발하여 HMM의 약점으로 지적되는 독립관측 가정을 완화시키고자 한다. 제안된 특징 표현은 단일 프레임 특징이 음성 신호의 시간적 동적 특성 (temporal dynamics)을 제대로 표현하지 못하기 때문에, 여러 프레임을 이용하여 음성 특징을 표현하도록 한다. 분절 특징은 다항식의 회귀 함수(polynomial regression function)에 의하여 관측 벡터의 궤적으로 표현되고, 이 특징을 패턴 분류에 사용하기 위하여 음성 신호의 궤적을 효과적으로 표현하는 분절 HMM(segmental HMM)을 이용한다. SHMM은 상태에서의 관측 확률을 외적 분절 변이와 내적 분절 변이로 세분하며, 외적 분절 변이는 장기적인 변화를, 내적 분절 변이는 단기적인 변화를 나타낸다. 음향학적 모델에서 분절 특성을 고려하기 위하여 외적 분절 변이는 분절의 확률 분포로 표현하고, 내적 분절 변이는 궤적의 추정 오차로 표현하도록 SHMM을 수정한 분절 특징 HMM(SFHMM; segmental-feature HMM)을 제안한다. SFHMM에서는 분절의 관측 확률을 분절 우도와 궤적의 추정 오차의 관계로써 표현하며, 추정오차는 특정 상태에서의 분절의 우도에 대한 가중치로 고려될 수 있다. 제안된 방법의 유효성과 분절 특징의 특성을 살펴보기 위하여 TIMIT 자료를 이용하여 몇 가지 실험을 하였다. 이들 실험 결과에서, 제안된 방법이 기존의 HMM보다 매개 변수가 많더라도, 성능의 향상과 제안된 특징이 유연하고 정보를 많이 가진다는 점에서 의미가 있다고 하겠다. In this paper, we propose a new acoustic model for characterizing segmental features and an algorithm based upon a general framework of hidden Markov models (HMMs) in order to compensate the weakness of HMM assumptions. The segmental features are represented as a trajectory of observed vector sequences by a polynomial regression function because the single frame feature cannot represent the temporal dynamics of speech signals effectively. To apply the segmental features to pattern classification, we adopted segmental HMM(SHMM) which is known as the effective method to represent the trend of speech signals. SHMM separates observation probability of the given state into extra- and intra-segmental variations that show the long-term and short-term variabilities, respectively. To consider the segmental characteristics in acoustic model, we present segmental-feature HMM(SFHMM) by modifying the SHMM. The SFHMM therefore represents the external- and internal-variation as the observation probability of the trajectory in a given state and trajectory estimation error for the given segment, respectively. We conducted several experiments on the TIMIT database to establish the effectiveness of the proposed method and the characteristics of the segmental features. From the experimental results, we conclude that the proposed method is valuable, if its number of parameters is greater than that of conventional HMM, in the flexible and informative feature representation and the performance improvement.

Removal of Heterogeneous Candidates Using Positional Accuracy Based on Levenshtein Distance on Isolated n-best Recognition

윤영선,Yun, Young-Sun The Acoustical Society of Korea 2011 韓國音響學會誌 Vol.30 No.8

Many isolated word recognition systems may generate irrelevant words for recognition results because they use only acoustic information or small amount of language information. In this paper, I propose word similarity that is used for selecting (or removing) less common words from candidates by applying Levenshtein distance. Word similarity is obtained by using positional accuracy that reflects the frequency information along to character's alignment information. This paper also discusses various improving techniques of selection of disparate words. The methods include different loss values, phone accuracy based on confusion information, weights of candidates by ranking order and partial comparisons. Through experiments, I found that the proposed methods are effective for removing heterogeneous words without loss of performance.

스마트 메모리 태그를 위한 Touch & Play 기반 저전력 소모 관리 기법

윤영선,하선주,손경아,은성배,Yun, Young-Sun,Ha, Sunju,Son, Kyung A,Eun, Seongbae 대한임베디드공학회 2017 대한임베디드공학회논문지 Vol.12 No.3

QR/NFC tags have been utilized in various fields like exhibition, museum, and so on, but they have a drawback that they are read-only fundamentally. We devise a novel device called a smart memory tag (below mem-tag) which is supplemented with the write function through combining a flash memory into a NFC tag. A mem-tag is composed of an NFC tag, an MCU, a flash memory, a bluetooth module, and a battery, and is inter-operating with smartphones via bluetooth communication. It can be used in a bulletin board to support writing replies and in a check-in service to verify the presence of the site. What matters is that users' inter-operations are borne to be asynchronous, which leads to the energy consumption to wait for users' actions. Sleep mechanisms and asynchronous MAC protocols used in ubiquotous sensor networks cannot avoid the consumption of battery. In this paper, we propose a touch and play scheme for minimizing the consumption of battery that the MCU wakes up and PLAY when a user TOUCH the mem-tag. We implemented the system to show that our scheme lets the mem-tag work 50 times longer than the sleep and wake-up scheme.

트리 구조 어휘 사전을 이용한 연결 숫자음 인식 시스템의 구현

윤영선,채의근,Yun Young-Sun,Chae Yi-Geun 대한음성학회 2004 말소리 Vol.50 No.-

In this paper, we consider the implementation of connected digit recognition system using tree structured lexicon model. To implement efficiently the fixed or variable length digit recognition system, finite state network (FSN) is required. We merge the word network algorithm that implements the FSN with lexical tree search algorithm that is used for general speech recognition system for fast search and large vocabulary systems. To find the efficient modeling of digit recognition system, we investigate some performance changes when the lexical tree search is applied.

성대위축증 및 경미한 성대구증에서 $Artecoll^{(R)}$을 이용한 후두주입성형술의 효과

윤영선,여진하,최지은,손영익,Yun, Young-Sun,Yeo, Jinha,Choi, Ji Eun,Son, Young-Ik 대한후두음성언어의학회 2013 대한후두음성언어의학회지 Vol.24 No.2

Background and Objectives : This study aimed to determine the clinical effect of $Artecoll^{(R)}$ injection laryngoplasty for patients with vocal atrophy and mild sulcus vocalis. Materials and Method : Forty-one patients with vocal atrophy and/or mild sulcus vocalis received transcutaneous $Artecoll^{(R)}$ injection into the vocal folds under local anesthesia. Subjective evaluations including voice handicap index (VHI) and perceptual grading with Grade, Roughness, Breathiness, Asthenia, Strain (GRBAS) scales and objective evaluations including jitter, shimmer, noise-to-harmonic ratio (NHR), speaking fundamental frequency (SFF) and maximum phonation time (MPT) were evaluated before and 3 months after the injection. Results : VHI and Grade, Breathiness and Strain scales in GRBAS showed significant improvement 3 months after injection. SFF and MPT also significantly improved after the injection ; MPT increased and SFF in male patients decreased. Conclusion : Injection laryngoplasty with $Artecoll^{(R)}$ is an effective method for correcting the glottal insufficiency and improving voice quality in patients with vocal atrophy and/or mild sulcus.

강인한 음성 인식을 위한 탠덤 구조와 분절 특징의 결합

윤영선,이윤근,Yun, Young-Sun,Lee, Yun-Keun 대한음성학회 2007 말소리 Vol.62 No.-

It is reported that the segmental feature based recognition system shows better results than conventional feature based system in the previous studies. On the other hand, the various studies of combining neural network and hidden Markov models within a single system are done with expectations that it may potentially combine the advantages of both systems. With the influence of these studies, tandem approach was presented to use neural network as the classifier and hidden Markov models as the decoder. In this paper, we applied the trend information of segmental features to tandem architecture and used posterior probabilities, which are the output of neural network, as inputs of recognition system. The experiments are performed on Auroral database to examine the potentiality of the trend feature based tandem architecture. From the results, the proposed system outperforms on very low SNR environments. Consequently, we argue that the trend information on tandem architecture can be additionally used for traditional MFCC features.

강인한 음성 인식을 위한 선형 로그 함수 기반의 MFCC 특징 표현 연구

윤영선,Yun, Young-Sun 대한음성학회 2006 말소리 Vol.59 No.-

In previous study, the linlog(linear log) RASTA(J-RASTA) approach based on PLP was proposed to deal with both the channel effect and the additive noise. The extraction of PLP required generally more steps and computation than the extraction of widely used MFCC. Thus, in this paper, we apply the linlog function to the MFCC for investigating the possibility of simple compensation method that removes both distortion. With the experimental results, the proposed method shows the similar tendency to the linlog RASTA-PLP_ When the J value is set to le-6, the best ERR(Error Reduction Rate) of 33% is obtained. For applying the linlog function to the feature extraction process, the J value plays a very important role in compensating the corruption. Thus, the study for the adaptive J or noise dependent J estimation is further required.

이웃 정보에 기초한 반모델을 이용한 발화 검증

윤영선,Yun, Young-Sun 대한음성학회 2008 말소리 Vol.67 No.-

In this paper, we investigate the relation between Bayes factor and likelihood ratio test (LRT) approaches and apply the neighborhood information of Bayes factor to building an alternate hypothesis model of the LRT system. To consider the neighborhood approaches, we contemplate a distance measure between models and algorithms to be applied. We also evaluate several methods to improve performance of utterance verification using neighborhood information. Among these methods, the system which adopts anti-models built by collecting mixtures of neighborhood models obtains maximum error rate reduction of 17% compared to the baseline, linear and weighted combination of neighborhood models.

분절특징 HMM의 특성에 관한 연구

윤영선,정호영,Yun Young-Sun,Jung Ho-Young 대한음성학회 2002 말소리 Vol.43 No.-

In this paper, we discuss the characteristics of Segmental-Feature HMM and summarize previous studies of SFHMM. There are several approaches to reduce the number of parameters in the previous studies. However, if the number of parameters decreased, the performance of systems also fell. Therefore, we consider the fast computation approach with preserving the same number of parameters. In this paper, we present the new segment comparison method to speed up the computation of SFHMM without loss of performance. The proposed method uses the three-frame calculation rather than the full(five) frames in the given segment. The experimental results show that the performance of the proposed system is better than that of the previous studies.

갑상선 수술 후 발생한 음성장애에 대한 음성치료 역할

윤영선,Yun, Young-Sun 대한후두음성언어의학회 2015 대한후두음성언어의학회지 Vol.26 No.1

Voice therapy after post-thyroidectomy dysphonia is designed to improve glottal closure without supraglottic hyperfunction by development of abdominal support for breathing and intrinsic muscle strengthening exercises. Regarding voice therapy of unilateral vocal fold paresis/paralysis, several techniques such as head positioning, digital manipulation, pushing and pulling method are used for trial or temporary therapy. And the holistic voice therapy such as LSVT, Accent method, VFE are useful for hypofunctional voice disorders. These may be effective interventions as a temporary improvement until normal voice recovers or may provide the essentials for long-term rehabilitation.