Publications – AISSRLAB

Music Rhythm Segmentation

JAN. 2022 – JUNE 2022

Article title:Tracking the Rhythm: Pansori Rhythm Segmentation and Classification Methods and Datasets.[SCIE Journal]
Music rhythm classification and segmetation dataset.
Korean traditional music (Pansori) dataset.
Classification and semantic segmentation methods for music rhythm segmentation.
HRnet, Novel optimized network and DeepLabV3 network architectures.
Puplished on Applied Sciences (MDPI)in Sep., 2022

Music Source Separation

OCT. 2021 – FEB. 2022

Article title:High-Resolution Representation Learning and Recurrent Neural Network for Singing Voice Separation.[SCIE Journal]
Music source separation using novel dataset.
Modified HRnet and ablation study.
Comparation on public dataset and state of art result.
Puplished on Circuits, Systems, and Signal Processing (Springer)in Sep., 2022

Music genre calssification

DEC. 2020 – SEP. 2021

Article title:Multi-modal, Multi-task and Multi-label for Music Genre Classification and Emotion Regression.[IEEE Conference]
Multimodal, multi-task and multi-label DNN.
Optimized neural network.
44 Class categores of music genra.
Puplished on ICTC2021in Dec. 2021

Plant Disease Classification

OCT. 2020 – MARCH 2021

Article title:An Incremental Learning for Plant Disease classification.[IEEE Conference]
Life-long learning method in DNN.
ResNet18 and ResNet50 neural network.
High performance.
Puplished on ICTC2021in Dec. 2021

Music Video Affective Computing (Unsupervised)

SEP. 2020 – JAN. 2021

Article title:Music video Emotion Classification using Slow-fast Audio-video Network and Unsupervised Feature Representation.[SCIE Journal]
Unsupervised and supervised music video emotion classification dataset
Autoencoder architecture with audio adn video information.
Slow-fast audio-video network to capture spatial and temporal information of music and video.
Train time information sharing and boosting modules.
Puplished on Scientific reports (nature)in Oct., 2021

Music Video Affective Computing (Supervised)

SEP. 2020 – JAN. 2021

Article title:Deep-Learning-Based Multimodal Emotion Classification for Music Videos.[SCIE Journal]
Music video emotion classification dataset (Inproved and Extended version)
Ablation study on unimodla and multimodal using music, video and facial expression.
Network complexity reduction using novel channel and filter separable convolution.
Train time information sharing and boosting modules.
End-to-end training, better result on visual and statistical analysis.
Puplished on Sensorsin July, 2021

Facemask States Detection

NOV. 2020 – JAN. 2021

Article title:Deep Learning Based Face Mask Status Detection for COVID-19.[Conference]
Semi-automatic visual object labeling tool
Facemask detection dataset with three cass categories of with mask, without mask and wrong weared mask.
Mask detetion using Faster-RCNN, Cascade FRCNN, FPN and Cascade FPN.
Comparision, visualization and analysis of sustem ability and applications.
Puplished on ICMLT Conferencein April 2021

Sound Event Labeling Tool

DEC. 2019 – SEP. 2020

Article title:A Semi-automatic Sound Annotation Tool for Audio/Video data.[SCIE Journal]
Semi-automatic sound event annotation tool using audio and video as input.
Automatic event detector is used to detect the audio event.
Based on the automatic detector result, an human annotation have to refine the annotation boundary.
Easy to use, better audio visualization, python based and output in easy CSV data file.
Diversified annotation tool for any rare sound event.
Puplished on Livestock Sciencein Feb. 2022

Music Source Seperation

JUNE 2020 – NOV. 2020

Article title:Parallel Stacked Hourglass Network for Music Source Separation.[SCIE Journal]
Prepared Korean traditional song (Pansori) dataset with 3 sources.
Korean traditional music Pansori dataset, MIR-1K dataset, and DSD100 dataset used in experiment.
Proposed a novel parallel stacked hourglass network (PSHN) with multiple band spectrograms.
Ablation study on proposed and past architecture.
State-of-art result.
Puplished on IEEE Accessin Nov. 2020

CNN Based Sound Event Detection in Cowshed

DEC. 2019 – SEP. 2020

Article title:Sound Event Detection in Cowshed using Synthetic data and Convolutional Neural Network[IEEE Conference]
CNN based sound event detection.
Sound event annotaion tool.
Sound localization and classification.
Puplished on ICTC2020in Sep. 2020

Cow Sound Event Localization and Classification

DEC. 2019 – SEP. 2020

Article title:Visual Object Detector for Cow Sound Event Detection[SCIE Journal]
Cow sound event detection dataset with 4 class categories.
CNN used for sound event detection using Cow sound dataset and UrbanSound8K dataset.
Visual object detection architecture (F-RCNN, CF-RCNN, FPN, C-FPC) used for audio event detection (in Log Mel-Spectrogram).
Compare the proposed CNN and Visual object detection architecture using three test dataset.
Puplished on IEEE Accessin Sep. 2020

Music-Video Emotion Classification

JAN. 2019 – SEP. 2019

Article title:Deep Learning-Based Late Fusion of Multimodal Information for Emotion Classification of Music Video[SCIE Journal]
Music-Video emotion classification using audio and video multimodal network architecture.
Use pretrained CNN for audio and 3D video model (I3D and C3D).
The network learned features ate late fused and compare the impact of network feature fusion.
Cross validation and network feature fusion.
Puplished on Multimedia Tools and Applicationsin Sep. 2020

Music Video Emotion Analysis

DEC. 2018 – MARCH 2019

Article title:Music-Video Emotion Analysis Using Late Fusion of Multimodal[Conference]
Music video emotion dataset of six class category.
Audio-video multimodal architecture.
C3D pretrained network and CNN pretrained audio network feature fusion.
Emotion representation in 2D emotion space.
Puplished on ITEEE 2019 Conferencein 2019

Domestic Cat Sound Classification

DEC. 2017 – SEP. 2018

Article title:Domestic Cat Sound Classification Using Learned Features from Deep Neural Nets[SCI Journal]
CNN and CDBN network architecture.
Cat sound dataset preparation of 10 class categories.
Frequency division average pooling (FDAP) technique instead of global average pooling (GAP) to make a robust prediction using various frequency band features.
Audio augmentation and learned feature visualization.
Puplished on Applied Sciencein Sep 2018

Domestic Cat Sound Classification using Transfer Learning

DEC. 2017 – MARCH 2018

Article title:Domestic Cat Sound Classification Using Transfer Learning[SCIE Journal]
Cat sound dataset with 10 class categories.
Use pretrained CNN for feature extraction and make feature classification.
Machine learning classifier and deep learning classifier comparision.
Ensemble and data augmentation.
Puplished on International Journal of Fuzzy Logic and Intelligent Systemsin June 2018

Current Research

Multi-culture emotion analysis.
Vegetable disease detection and remedy system.
Remote sensing and GIS image analysis for agriculture field monitoring.
Wildlife activity monitoring.
Authentication system for vahicle.

Artificial Intelligence and Smart System Research Lab

Artificial Intelligence and
Smart System Research Lab

Music Rhythm Segmentation

Music Source Separation

Music genre calssification

Plant Disease Classification

Music Video Affective Computing (Unsupervised)

Music Video Affective Computing (Supervised)

Facemask States Detection

Sound Event Labeling Tool

Music Source Seperation

CNN Based Sound Event Detection in Cowshed

Cow Sound Event Localization and Classification

Music-Video Emotion Classification

Music Video Emotion Analysis

Domestic Cat Sound Classification

Domestic Cat Sound Classification using Transfer Learning

Current Research

Links

Contact Info

Subscribe to newsletter