Pub Date : 2022-07-06DOI: 10.1109/ICCE-Taiwan55306.2022.9869188
Yaozong Gan, Ren Togo, Takahiro Ogawa, M. Haseyama
This paper presents a scene retrieval method in soccer videos with video vision Transformer (ViViT). In soccer coaching, it is difficult for the training staff to find the required scenes efficiently from the large number of soccer videos. We tackle this problem with a simple yet effective method. We train ViViT and obtain the output token features of the soccer scene by the pre-trained ViViT model. The output tokens of the pre-trained ViViT contain spatio-temporal information of soccer scenes. We then transform a query scene and candidate scenes into output token features using the pre-trained ViViT and calculate the similarity between the tokens with cosine similarity. We conducted experiments on SoccerNet-V2dataset. The experimental results show that the proposed method achieves outstanding retrieval accuracy compared to the previous methods.
{"title":"Scene Retrieval in Soccer Videos by Spatial-temporal Attention with Video Vision Transformer","authors":"Yaozong Gan, Ren Togo, Takahiro Ogawa, M. Haseyama","doi":"10.1109/ICCE-Taiwan55306.2022.9869188","DOIUrl":"https://doi.org/10.1109/ICCE-Taiwan55306.2022.9869188","url":null,"abstract":"This paper presents a scene retrieval method in soccer videos with video vision Transformer (ViViT). In soccer coaching, it is difficult for the training staff to find the required scenes efficiently from the large number of soccer videos. We tackle this problem with a simple yet effective method. We train ViViT and obtain the output token features of the soccer scene by the pre-trained ViViT model. The output tokens of the pre-trained ViViT contain spatio-temporal information of soccer scenes. We then transform a query scene and candidate scenes into output token features using the pre-trained ViViT and calculate the similarity between the tokens with cosine similarity. We conducted experiments on SoccerNet-V2dataset. The experimental results show that the proposed method achieves outstanding retrieval accuracy compared to the previous methods.","PeriodicalId":164671,"journal":{"name":"2022 IEEE International Conference on Consumer Electronics - Taiwan","volume":"78 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116078375","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-07-06DOI: 10.1109/ICCE-Taiwan55306.2022.9869081
Yufeng Li, Lin Shang, Peng Pan
This paper studies the abnormal posture detection algorithm based on heuristic optimization. Using the data collected by sensors, the features such as acceleration and angular velocity are extracted and put into the classifiers for training. We select the appropriate heuristic algorithm according to different classifier models for optimization. The results demonstrate that, in binary classification experiment, the accuracy ratio of the K-Nearest Neighbor (KNN) model is 99.54%, and the AUC is 0.99. In quad classification experiment, The Support Vector Machine (SVM) model has a 94.32% accuracy ratio and a 0.95 AUC, which has the optimal performance.
{"title":"Heuristic Optimization based Abnormal Posture Detection Algorithm","authors":"Yufeng Li, Lin Shang, Peng Pan","doi":"10.1109/ICCE-Taiwan55306.2022.9869081","DOIUrl":"https://doi.org/10.1109/ICCE-Taiwan55306.2022.9869081","url":null,"abstract":"This paper studies the abnormal posture detection algorithm based on heuristic optimization. Using the data collected by sensors, the features such as acceleration and angular velocity are extracted and put into the classifiers for training. We select the appropriate heuristic algorithm according to different classifier models for optimization. The results demonstrate that, in binary classification experiment, the accuracy ratio of the K-Nearest Neighbor (KNN) model is 99.54%, and the AUC is 0.99. In quad classification experiment, The Support Vector Machine (SVM) model has a 94.32% accuracy ratio and a 0.95 AUC, which has the optimal performance.","PeriodicalId":164671,"journal":{"name":"2022 IEEE International Conference on Consumer Electronics - Taiwan","volume":"24 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114639628","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-07-06DOI: 10.1109/ICCE-Taiwan55306.2022.9869211
Tsung-Han Tsai, Ping-Cheng Hao, Fong-Lin Tsai
In this paper, a customized speaker verification system combined with noise-cancellation using blind source separation was proposed. This system is divided into two phases: the noise-cancellation phase and the speaker verification phase. In the noise-cancellation phase, a fast time-frequency mask technique based on Short Time Fourier Transform (STFT) was proposed for separating a mixture of two input sounds in a single signal. After obtaining the separated speech data, this input is processed to the wake-up word system. In the speaker verification phase, we use Mel-Frequency Cepstral Coefficients (MFCC) as the feature extraction module. Then we train the feature data into a voiceprint model and a state sequence model of the speaker using Gaussian mixture model (GMM) and hidden Markov model (HMM), respectively. An analysis is done on noisy speech signals corrupted by white noise at different angles. Based on the output SIR (Signal to Interference Ratio) and SDR (Signal to Distortion Ratio) analysis, the improved accuracy is derived in the proposed system. We have obtained promising results in the real experimental environment.
{"title":"Customized Speaker Verification System with Noise-Cancellation using Blind Source Separation","authors":"Tsung-Han Tsai, Ping-Cheng Hao, Fong-Lin Tsai","doi":"10.1109/ICCE-Taiwan55306.2022.9869211","DOIUrl":"https://doi.org/10.1109/ICCE-Taiwan55306.2022.9869211","url":null,"abstract":"In this paper, a customized speaker verification system combined with noise-cancellation using blind source separation was proposed. This system is divided into two phases: the noise-cancellation phase and the speaker verification phase. In the noise-cancellation phase, a fast time-frequency mask technique based on Short Time Fourier Transform (STFT) was proposed for separating a mixture of two input sounds in a single signal. After obtaining the separated speech data, this input is processed to the wake-up word system. In the speaker verification phase, we use Mel-Frequency Cepstral Coefficients (MFCC) as the feature extraction module. Then we train the feature data into a voiceprint model and a state sequence model of the speaker using Gaussian mixture model (GMM) and hidden Markov model (HMM), respectively. An analysis is done on noisy speech signals corrupted by white noise at different angles. Based on the output SIR (Signal to Interference Ratio) and SDR (Signal to Distortion Ratio) analysis, the improved accuracy is derived in the proposed system. We have obtained promising results in the real experimental environment.","PeriodicalId":164671,"journal":{"name":"2022 IEEE International Conference on Consumer Electronics - Taiwan","volume":"67 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114768232","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The application of artificial intelligence to the diagnosis of ECG is of great significance. We combine machine learning algorithm with deep learning algorithm to give full play to the advantages of different algorithms by ensemble learning. Finally, we fuse the selected models so that the accuracy of identifying five kinds of arrhythmias can reach 94%. Particularly, the accuracy of class F beat which is difficult to identify has also been improved.
{"title":"An Integration Method for ECG Multi-Classification","authors":"Chao-Xin Xie, Minghui Fan, Liang-Hung Wang, Pao-Cheng Huang","doi":"10.1109/ICCE-Taiwan55306.2022.9869291","DOIUrl":"https://doi.org/10.1109/ICCE-Taiwan55306.2022.9869291","url":null,"abstract":"The application of artificial intelligence to the diagnosis of ECG is of great significance. We combine machine learning algorithm with deep learning algorithm to give full play to the advantages of different algorithms by ensemble learning. Finally, we fuse the selected models so that the accuracy of identifying five kinds of arrhythmias can reach 94%. Particularly, the accuracy of class F beat which is difficult to identify has also been improved.","PeriodicalId":164671,"journal":{"name":"2022 IEEE International Conference on Consumer Electronics - Taiwan","volume":"14 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123681849","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-07-06DOI: 10.1109/ICCE-Taiwan55306.2022.9869000
Kam-Hong Chan, C. Chao
The identification of drivers is essential for many applications, such as attribution of liability for car accidents and driving risk assessment. Most existing driver identification systems adopt identity keys (such as car keys and smart cards) or biometrics technology (such as face recognition, iris recognition, fingerprint recognition, voiceprint recognition, and vein recognition, etc.) to identify drivers. However, these schemes are unable to detect driver changes during a trip. In this paper, combining voiceprint and acoustic driving characteristics, the driver identity system DriverID is proposed to identify the person who is actually driving. DriverID uses the Deep Residual Network (ResNet) to construct an acoustic recognition model based on the voice key recorded by the driver. In addition, the Convolutional Neural Network (CNN) is used to construct an acoustic driving action recognition model based on the reflection of acoustic signals generated by the user. Combining the two recognition methods, DriverID can correctly identify the driver with high probability. It is believed that DriverID is a practical driver identity system.
{"title":"DriverID: Driver Identity System Based on Voiceprint and Acoustic Sensing","authors":"Kam-Hong Chan, C. Chao","doi":"10.1109/ICCE-Taiwan55306.2022.9869000","DOIUrl":"https://doi.org/10.1109/ICCE-Taiwan55306.2022.9869000","url":null,"abstract":"The identification of drivers is essential for many applications, such as attribution of liability for car accidents and driving risk assessment. Most existing driver identification systems adopt identity keys (such as car keys and smart cards) or biometrics technology (such as face recognition, iris recognition, fingerprint recognition, voiceprint recognition, and vein recognition, etc.) to identify drivers. However, these schemes are unable to detect driver changes during a trip. In this paper, combining voiceprint and acoustic driving characteristics, the driver identity system DriverID is proposed to identify the person who is actually driving. DriverID uses the Deep Residual Network (ResNet) to construct an acoustic recognition model based on the voice key recorded by the driver. In addition, the Convolutional Neural Network (CNN) is used to construct an acoustic driving action recognition model based on the reflection of acoustic signals generated by the user. Combining the two recognition methods, DriverID can correctly identify the driver with high probability. It is believed that DriverID is a practical driver identity system.","PeriodicalId":164671,"journal":{"name":"2022 IEEE International Conference on Consumer Electronics - Taiwan","volume":"50 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123907479","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-07-06DOI: 10.1109/ICCE-Taiwan55306.2022.9869122
Kohei Tsunewaki, Tomotaka Kimura, Jun Cheng
In this paper, we propose a ransomware detection method based on API IDs and call intervals. In the proposed method, to detect ransomware, when each API call occurs, we input both the API ID and the call interval from the previous call into an LSTM (Long Short Term Memory). By inputting the API IDs and call intervals into LSTM, we can learn the characteristics of the time series change of API calls in the ransomware. Through the experiments using an original dataset, we demonstrated that the accuracy of our proposed method was high and the characteristic learning of the call interval was useful for detecting ransomware.
{"title":"LSTM-Based Ransomware Detection Using API Call Information","authors":"Kohei Tsunewaki, Tomotaka Kimura, Jun Cheng","doi":"10.1109/ICCE-Taiwan55306.2022.9869122","DOIUrl":"https://doi.org/10.1109/ICCE-Taiwan55306.2022.9869122","url":null,"abstract":"In this paper, we propose a ransomware detection method based on API IDs and call intervals. In the proposed method, to detect ransomware, when each API call occurs, we input both the API ID and the call interval from the previous call into an LSTM (Long Short Term Memory). By inputting the API IDs and call intervals into LSTM, we can learn the characteristics of the time series change of API calls in the ransomware. Through the experiments using an original dataset, we demonstrated that the accuracy of our proposed method was high and the characteristic learning of the call interval was useful for detecting ransomware.","PeriodicalId":164671,"journal":{"name":"2022 IEEE International Conference on Consumer Electronics - Taiwan","volume":"43 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124156765","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-07-06DOI: 10.1109/ICCE-Taiwan55306.2022.9869210
Kohei Hayashi, Rrota Horiuchi, N. Komuro
In recent years, WSNs have been expected to be applied to various fields such as home security, healthcare, and environmental monitoring. A number of studies have been done on IEEE 802.15.4, the standard for WSNs, but they have left some problems and there is still room for improvement. In this paper, we propose a new access control protocol in tree-type heterogeneous sensor networks that achieves low EC, high PDR, and low latency by adjusting the active period so that the buffer occupancy ratio of the relay node is less than 1 to prevent the buffer from overflowing, and then performing channel partitioning and scheduling to avoid packet collisions.
{"title":"Scheduling Method for Improving Transmission and Reception Efficiency in IEEE802.15.4 used Heterogeneous Wireless Sensor Networks","authors":"Kohei Hayashi, Rrota Horiuchi, N. Komuro","doi":"10.1109/ICCE-Taiwan55306.2022.9869210","DOIUrl":"https://doi.org/10.1109/ICCE-Taiwan55306.2022.9869210","url":null,"abstract":"In recent years, WSNs have been expected to be applied to various fields such as home security, healthcare, and environmental monitoring. A number of studies have been done on IEEE 802.15.4, the standard for WSNs, but they have left some problems and there is still room for improvement. In this paper, we propose a new access control protocol in tree-type heterogeneous sensor networks that achieves low EC, high PDR, and low latency by adjusting the active period so that the buffer occupancy ratio of the relay node is less than 1 to prevent the buffer from overflowing, and then performing channel partitioning and scheduling to avoid packet collisions.","PeriodicalId":164671,"journal":{"name":"2022 IEEE International Conference on Consumer Electronics - Taiwan","volume":"85 11 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122666284","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-07-06DOI: 10.1109/ICCE-Taiwan55306.2022.9869273
Kuan-Ting Lai, Guo-Shiang Lin
In this paper, we propose a light-weight convolutional neural network based on depth-wise separable convolutions and cross stage partial (CSP) network. Dissimilar to MobileNetV3, the proposed network is composed of some CSP blocks to reduce the model size and computational operations. For performance evaluation, Cifar10 and Cifar100 are used for testing. Compared to MobileNetv3, the model size and execution time of the proposed network in PC and mobile device are smaller. Therefore, the experimental results show that the proposed light-weight network can effectively extract visual features for image classification compared with MobileNetV3.
{"title":"A New Light Weight Convolutional Neural Network for Mobile Devices","authors":"Kuan-Ting Lai, Guo-Shiang Lin","doi":"10.1109/ICCE-Taiwan55306.2022.9869273","DOIUrl":"https://doi.org/10.1109/ICCE-Taiwan55306.2022.9869273","url":null,"abstract":"In this paper, we propose a light-weight convolutional neural network based on depth-wise separable convolutions and cross stage partial (CSP) network. Dissimilar to MobileNetV3, the proposed network is composed of some CSP blocks to reduce the model size and computational operations. For performance evaluation, Cifar10 and Cifar100 are used for testing. Compared to MobileNetv3, the model size and execution time of the proposed network in PC and mobile device are smaller. Therefore, the experimental results show that the proposed light-weight network can effectively extract visual features for image classification compared with MobileNetV3.","PeriodicalId":164671,"journal":{"name":"2022 IEEE International Conference on Consumer Electronics - Taiwan","volume":"74 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128217664","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-07-06DOI: 10.1109/ICCE-Taiwan55306.2022.9868978
Sue-Yi Chen, Shao-Chang Huang, K. Hsu, Yin-Wei Peng, Jiabin Dong, J. Gan
Large substrate currents could induce the device melt in power manager integrated circuit applications. Many researches are focused on how to reduce substrate currents from process modifications. In this paper, the fundamental substrate current mechanism analyses are studied. Then, a tracing-high voltage between the device drain terminal and the device source terminal applied on the isolation ring is proposed for substrate current reductions. Engineers can apply this method for avoiding the device burned-out without the complicated process changes.
{"title":"Reducing Large LDNMOSFET Substrate Currents by Modifying Isolation Ring Voltages","authors":"Sue-Yi Chen, Shao-Chang Huang, K. Hsu, Yin-Wei Peng, Jiabin Dong, J. Gan","doi":"10.1109/ICCE-Taiwan55306.2022.9868978","DOIUrl":"https://doi.org/10.1109/ICCE-Taiwan55306.2022.9868978","url":null,"abstract":"Large substrate currents could induce the device melt in power manager integrated circuit applications. Many researches are focused on how to reduce substrate currents from process modifications. In this paper, the fundamental substrate current mechanism analyses are studied. Then, a tracing-high voltage between the device drain terminal and the device source terminal applied on the isolation ring is proposed for substrate current reductions. Engineers can apply this method for avoiding the device burned-out without the complicated process changes.","PeriodicalId":164671,"journal":{"name":"2022 IEEE International Conference on Consumer Electronics - Taiwan","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128939106","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-07-06DOI: 10.1109/ICCE-Taiwan55306.2022.9869279
Jian-Chiuan Hou, Weimin Xu, Yuanyuan Chu, Chih-Lin Hu, Ying-Hong Chen, Shi Chen, Lin Hui
We propose a fall detection mechanism based on multi-camera cooperation in home space. Cameras capture image-based falling events, and self-organize a group using deep reinforcement learning. Neighbor cameras exchange sensing data and statuses in local network proximity. With information sharing in a group, cameras can improve the accuracy of decision making on falling events and cope with the limited fields of view against physical deployment of cameras in residential areas.
{"title":"Cooperative Fall Detection with Multiple Cameras","authors":"Jian-Chiuan Hou, Weimin Xu, Yuanyuan Chu, Chih-Lin Hu, Ying-Hong Chen, Shi Chen, Lin Hui","doi":"10.1109/ICCE-Taiwan55306.2022.9869279","DOIUrl":"https://doi.org/10.1109/ICCE-Taiwan55306.2022.9869279","url":null,"abstract":"We propose a fall detection mechanism based on multi-camera cooperation in home space. Cameras capture image-based falling events, and self-organize a group using deep reinforcement learning. Neighbor cameras exchange sensing data and statuses in local network proximity. With information sharing in a group, cameras can improve the accuracy of decision making on falling events and cope with the limited fields of view against physical deployment of cameras in residential areas.","PeriodicalId":164671,"journal":{"name":"2022 IEEE International Conference on Consumer Electronics - Taiwan","volume":"18 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128812901","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：