Sunghyun Kim

Sunghyun Kim

Auto-generated profile, based on publicly available information

Sunghyun Kim

School of Robotics, Kwangwoon University, Seoul, South Korea

Research Focus & Metrics

Academic Career Length

5Years (2022-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
2 March 2022 in IEEE Access
Verified
Journal Article
Open access
SMSungwoo Moon
Sungwoo Moon
SKSunghyun Kim
Sunghyun Kim
YCYong-Hoon Choi
Yong-Hoon Choi
With the development of voice synthesis technology using deep learning, voice synthesis research that expresses the characteristics and emotions of speakers is actively being conducted. Current technology does not satisfactorily express various emotions and characteristics for speakers with very low or high vocal ranges and for speakers with dialects. In this paper, we propose mel-spectrogram image transfer (MIST)-Tacotron, a Tacotron 2-based speech synthesis model that adds a reference encoder with an image style transfer module. The proposed method is a technique that adds image style transfer to the existing Tacotron 2 model and extracts the speaker’s feature from the reference mel-spectrogram using a pre-trained deep learning model. Through the extracted feature, the style such as pitch, tone, and duration of the speaker are trained to express the style and emotion of the speaker more clearly. To extract the speaker’s style independently from the speaker’s timbre and emotion, the ID value for the speaker and the ID value for the emotional state were used as inputs. Performance is evaluated by F0 voiced error (FVE), F0 gross pitch error (F0 GPE), mel-cepstral distortion (MCD), band aperiodicity distortion (BAPD), voiced/unvoiced error (VUVE), false positive rate (FPR), and false negative rate (FNR). The performance of the proposed model was observed to have lower error values than the existing models, GST (Global Style Token) Tacotron and VAE (Variational Autoencoder) Tacotron. As a result of measuring mean opinion score (MOS), the sound quality of the proposed model received the highest score in terms of emotional expression and speaker style reflection.

4 February 2022 in IEEE Access
Verified
Journal Article
Open access
DKDaegyeom Kim
Daegyeom Kim
MKMyeongjin Ko
Myeongjin Ko
SKSunghyun Kim
Sunghyun Kim
SMSungwoo Moon
Sungwoo Moon
KCKyung-Yul Cheon
Kyung-Yul Cheon
YCYong-Hoon Choi
Yong-Hoon Choi
This paper proposes a neural 5G traffic generation model and a methodology for calculating the spectrum requirements of private 5G networks to provide various industrial communication services. To accurately calculate the spectral requirements, it is necessary to analyze the actual data volume and traffic type of industrial cases. However, because there is currently no suitable traffic model to test loads in private 5G networks, we have developed a generative adversarial network (GAN)-based traffic generator that can generate realistic traffic by learning actual traffic traces collected by mobile network operators. In addition, in the case of industrial applications, probability-based traffic models were used in parallel as there were not enough real data to be learned. The proposed 5G traffic generation model is combined with the proposed 5G spectrum calculation methodology, enabling more accurate spectrum requirements calculation through traffic simulation similar to a real-life environment. In this paper, the spectrum requirements are calculated differently according to two types of duplexing, namely frequency division duplexing (FDD) and time division duplexing (TDD). As a guide for companies aiming to provide advanced wireless connectivity for a wide variety of vertical industries using 5G networks, eight use cases defined in the 5G Alliance for Connected Industries and Automation (ACIA) white paper were simulated. The spectrum requirements were calculated under various simulation conditions considering varying traffic loads, deployment scenarios, and duplexing types. Various simulation results confirmed that a bandwidth of at least 22.0 MHz to a maximum of 397.8 MHz is required depending on the deployment scenario.

23 May 2022 in arXiv
Verified
Preprint
Open access
PDF file
SPSeo Jin Park
Seo Jin Park
JFJoshua Fried
Joshua Fried
SKSunghyun Kim
Sunghyun Kim
MAMohammad Alizadeh
Mohammad Alizadeh
ABAdam Belay
Adam Belay
As emerging deep neural network (DNN) models continue to grow in size, using large GPU clusters to train DNNs is becoming an essential requirement to achieving acceptable training times. In this paper, we consider the case where future increases in cluster size will cause the global batch size that can be used to train models to reach a fundamental limit: beyond a certain point, larger global batch sizes cause sample efficiency to degrade, increasing overall time to accuracy. As a result, to achieve further improvements in training performance, we must instead consider "strong scaling" strategies that hold the global batch size constant and allocate smaller batches to each GPU. Unfortunately, this makes it significantly more difficult to use cluster resources efficiently. We present DeepPool, a system that addresses this efficiency challenge through two key ideas. First, burst parallelism allocates large numbers of GPUs to foreground jobs in bursts to exploit the unevenness in parallelism across layers. Second, GPU multiplexing prioritizes throughput for foreground training jobs, while packing in background training jobs to reclaim underutilized GPU resources, thereby improving cluster-wide utilization. Together, these two ideas enable DeepPool to deliver a 1.2 - 2.3x improvement in total cluster throughput over standard data parallelism with a single task when the cluster scale is large.