Dance motion generation by recombination of body parts from motion source

Lee, Minho,Lee, Kyogu,Lee, Mihee,Park, Jaeheung Springer-Verlag 2018 INTELLIGENT SERVICE ROBOTICS Vol.11 No.2

<P>In this paper, we propose an approach to synthesize new dance routines by combining body part motions from a human motion database. The proposed approach aims to provide a movement source to allow robots or animation characters to perform improvised dances to music, and also to inspire choreographers with the provided movements. Based on the observation that some body parts perform more appropriately than other body parts during dance performances, a correlation analysis of music and motion is conducted to identify the expressive body parts. We then combine the body part movement sources to create a new motion, which differs from all sources in the database. The generated performances are evaluated by a user questionnaire assessment, and the results are discussed to understand what is important in generating more appealing dance routines.</P>

Using Experts Among Users for Novel Movie Recommendations

Lee, Kibeom,Lee, Kyogu Korean Institute of Information Scientists and Eng 2013 Journal of Computing Science and Engineering Vol.7 No.1

The introduction of recommender systems to existing online services is now practically inevitable, with the increasing number of items and users on online services. Popular recommender systems have successfully implemented satisfactory systems, which are usually based on collaborative filtering. However, collaborative filtering-based recommenders suffer from well-known problems, such as popularity bias, and the cold-start problem. In this paper, we propose an innovative collaborative-filtering based recommender system, which uses the concepts of Experts and Novices to create fine-grained recommendations that focus on being novel, while being kept relevant. Experts and Novices are defined using pre-made clusters of similar items, and the distribution of users' ratings among these clusters. Thus, in order to generate recommendations, the experts are found dynamically depending on the seed items of the novice. The proposed recommender system was built using the MovieLens 1 M dataset, and evaluated with novelty metrics. Results show that the proposed system outperforms matrix factorization methods according to discovery-based novelty metrics, and can be a solution to popularity bias and the cold-start problem, while still retaining collaborative filtering.

Using Dynamically Promoted Experts for Music Recommendation

Lee, Kibeom,Lee, Kyogu IEEE 2014 IEEE transactions on multimedia Vol.16 No.5

<P>Recommender systems have become an invaluable asset to online services with the ever-growing number of items and users. Most systems focused on recommendation accuracy, predicting likable items for each user. Such methods tend to generate popular and safe recommendations, but fail to introduce users to potentially risky, yet novel items that could help in increasing the variety of items consumed by the users. This is known as popularity bias, which is predominant in methods that adopt collaborative filtering. Recently, however, recommenders have started to improve their methods to generate lists that encompass diverse items that are both accurate and novel through specific novelty-driven algorithms or hybrid recommender systems. In this paper, we propose a recommender system that uses the concepts of Experts to find both novel and relevant recommendations. By analyzing the ratings of the users, the algorithm promotes special Experts from the user population to create novel recommendations for a target user. Thus, different users are promoted dynamically to Experts depending on who the recommendations are for. The system used data collected from Last.fm and was evaluated with several metrics. Results show that the proposed system outperforms matrix factorization methods in finding novel items and performs on par in finding simultaneously novel and relevant items. This system can also provide a means to popularity bias while preserving the advantages of collaborative filtering.</P>

Improving Diversity of Radio Playlists with the Aid of Recommender Systems

Kibeom Lee(이기범),Ziwon Hyung(형지원),Namjun Kang(강남준),Kyogu Lee(이교구) 한국방송학회 2014 한국방송학회 학술대회 논문집 Vol.2014 No.4

Using Experts Among Users for Novel Movie Recommendations

Kibeom Lee,Kyogu Lee 한국정보과학회 2013 Journal of Computing Science and Engineering Vol.7 No.1

The introduction of recommender systems to existing online services is now practically inevitable, with the increasing number of items and users on online services. Popular recommender systems have successfully implemented satisfactory systems, which are usually based on collaborative filtering. However, collaborative filtering-based recommenders suffer from well-known problems, such as popularity bias, and the cold-start problem. In this paper, we propose an innovative collaborative-filtering based recommender system, which uses the concepts of Experts and Novices to create finegrained recommendations that focus on being novel, while being kept relevant. Experts and Novices are defined using pre-made clusters of similar items, and the distribution of users’ ratings among these clusters. Thus, in order to generate recommendations, the experts are found dynamically depending on the seed items of the novice. The proposed recommender system was built using the MovieLens 1 M dataset, and evaluated with novelty metrics. Results show that the proposed system outperforms matrix factorization methods according to discovery-based novelty metrics, and can be a solution to popularity bias and the cold-start problem, while still retaining collaborative filtering.

음악에서의 보컬 신호 인식을 위한 트레이닝 데이터 자동주석기법 연구

이교구(Kyogu Lee),Markus Cremer 한국전자음악협회 2009 에밀레 Vol.- No.7

우리는 음악 신호에서 보컬과(vocal) 비보컬(non-vocal) 신호를 분리하기 위하여 최소한의 노동력으로 많은 양의 트레이닝 데이터를 레이블(label)하는 새로운 방법을 제시한다. 이를 위하여 보컬이 분리된 채널에 인코딩되어 있는 미디(MIDI) 파일을 합성함으로써 오디오 파일을 생성한 후, 합성된 오디오를 다이나믹 타임 워핑(Dynamic Time Warping) 알고리즘을 이용하여 실제 오디오와 정렬한다. 미디 파일의 보컬 라인에 포함되어 있는 노트 온/오프 정보는 정확한 보컬/비보컬 경계를 제공하고, 최소비용 정렬 궤도로부터 실제 리코딩에서도 상응하는 경계를 구한다. 이와 같이 노동력으로부터 자유로운 레이블링 과정을 이용하여 대규모의 트레이닝 데이터를 구축할 수 있으며, 히든 마르코프 모델을 인식기로 할 경우 기대되는 결과를 얻을 수 있음을 실험을 통하여 보인다. 또한 데이터의 규모가 증가함에 따라 성능도 향상되는 것을 보여줌으로써, 제안된 방법을 통하여 생성된 데이터의 유용성을 입증한다. We present a novel approach to labeling a large amount of training data for vocal/non-vocal discrimination in musical audio with the minimum amount of human labor. To this end, we use MIDI files for which vocal lines are encoded on a separate channel and synthesize them to create audio files. We then align synthesized audio with real recordings using dynamic time warping(DTW) algorithm. Note onset/offset information encoded in vocal lines in MIDI files provides precise vocal/non-vocal boundaries and we obtain from the minimum-cost alignment path the corresponding boundaries in actual recordings. This near labor-free labeling process allows us to acquire a large training data set, and the experiments show promising results when tested on an independent test set, using hidden Markov models as a classifier. We also demonstrate that the data generated by the proposed system is good data by showing that the overall performance increases with more training data.

오디오신호의 음계 변조를 이용한 음악 스타일 변환 방법

이장우(Jangwoo Lee),김현우(Hyunwoo Kim),이교구(Kyogu Lee) 한국전자음악협회 2010 에밀레 Vol.- No.8

음악의 조성(調性)은 음악의 성격이나 작곡자나 연주자가 전달하고자 하는 바를 전달함에 있어서 중요한 비중을 차지한다. 장조(長調)와 단조(短調)에서 특징적으로 두드러지는 구성음을 바꾸는 것은 음계를 바꾸게 된다. 그러나 구성음을 바꾸기 위해서는 그에 상응하는 배음(倍音)들 또한 변환이 요구된다. 이에 본 연구에서는 장조음계와 단조음계의 구성음을 웨이브 신호에서도 변환할 수 있는 방법을 제안한다. 우선 MIDI로 이루어진 동일한 곡을 장조와 단조로 만든 후 웨이브 시그널의 형태로 변환한다. 그리고 이것들을 학습 데이터로 이용하여, 상대 음계의 주파수 스펙트럼으로 변환할 수 있는 변환 행렬을 구한다. 마지막으로 학습되지 않은 데이터 즉, 새로운 장조 혹은 단조의 곡을 변환 행렬을 통해 새로운 음계의 곡을 생성함을 보인다. In tonal music, mode is concept that involves musical scales and melody type. So the musical scale of songs determines the style of them. In this paper, we change the musical scale of a song to change the style of a song. Without sinusoidal modeling, this method directly changes the energy of spectrum into the counterpart mode scale using transformation function. In order to calculate transformation matrix, we use synthesized acoustic training data based on midifiles. Transformation function is obtained by multiplying a pseudo inverse matrix of a spectrum and the counterpart. This is a promising algorithm to develop automatic arrangement system on acoustic tracks.

추체외로 증상에 따른 항정신병 약물 복용량과 음성 특성의 상관관계 분석

이수빈,김서영,김혜윤,김의태,유경상,이호영,이교구,Lee, Subin,Kim, Seoyoung,Kim, Hye Yoon,Kim, Euitae,Yu, Kyung-Sang,Lee, Ho-Young,Lee, Kyogu 한국음향학회 2022 韓國音響學會誌 Vol.41 No.3

본 논문은 항정신병 약물의 복용량에 따른 음성 특징의 상관관계 분석을 수행하였다. 항정신병 약물의 대표적 부작용 중 하나인 추체외로 증상(ExtraPyramidal Symptoms, EPS) 발생에 따른 음성 특징의 패턴을 알아보기 위하여, 문장 개발을 통해 한국어 기반 추체외로 증상 음성 코퍼스를 구축하였다. 수집된 자료는 추체외로 증상 군과 비 추체외로 증상 군으로 나누어 음성 특징 패턴을 조사하였으며, 특히 추체외로 증상 군의 높은 음성 특징 상관관계를 보였다. 또한, 발화 문장의 종류가 음성 특징 패턴에 영향을 미친다는 것을 확인할 수 있었으며, 이를 통해 음성 특징을 기반한 추체외로 증상의 조기 발견 가능성을 기대해볼 수 있었다.

Minimum Cost Data Aggregation for Wireless Sensor Networks Computing Functions of Sensed Data

Chen, Chao,Lee, Kyogu,Park, Joon-Sang,Baek, Seung Jun Hindawi Limited 2015 Journal of sensors Vol.2015 No.-

<P>We consider a problem of minimum cost (energy) data aggregation in wireless sensor networks computing certain functions of sensed data. We use in-network aggregation such that data can be combined at the intermediate nodes en route to the sink. We consider two types of functions: firstly the summation-type which includes<I>sum</I>,<I>mean</I>, and<I>weighted sum</I>, and secondly the extreme-type which includes<I>max</I>and<I>min</I>. However for both types of functions the problem turns out to be NP-hard. We first show that, for<I>sum</I>and<I>mean</I>, there exist algorithms which can approximate the optimal cost by a factor logarithmic in the number of sources. For<I>weighted sum</I>we obtain a similar result for Gaussian sources. Next we reveal that the problem for extreme-type functions is intrinsically different from that for summation-type functions. We then propose a novel algorithm based on the crucial tradeoff in reducing costs between local aggregation of flows and finding a low cost path to the sink: the algorithm is shown to empirically find the best tradeoff point. We argue that the algorithm is applicable to many other similar types of problems. Simulation results show that significant cost savings can be achieved by the proposed algorithm.</P>

Detecting fingering of overblown flute sound using sparse feature learning

Han, Yoonchang,Lee, Kyogu Springer (Biomed Central Ltd.) 2016 EURASIP journal on audio, speech, and music proces Vol.2016 No.-