Cheng-Hsiung Lee - Academia.edu (original) (raw)

Papers by Cheng-Hsiung Lee

This paper presents a new method to automatically grade pathological prostate images according to... more This paper presents a new method to automatically grade pathological prostate images according to Gleason grading system. Two feature extraction methods were proposed based on fractal dimension to analyze the variations of intensity and texture complexity in images. Each image can be classified into appropriate grade by using Bayes classifier and k-Nearest-Neighbor (k-NN) classifier, respectively. Leaving-One-Out approach was used to estimate the correct classification rates. Experimental results showed that 92.86 % of accuracy can be achieved by using Bayes classifier and 89.01 % of accuracy can be achieved by using k-NN classifier for a set of 182 pathological prostate images.

Computational Intelligence and Neuroscience, 2020

Previous studies have shown that training a reinforcement model for the sorting problem takes ver... more Previous studies have shown that training a reinforcement model for the sorting problem takes very long time, even for small sets of data. To study whether transfer learning could improve the training process of reinforcement learning, we employ Q-learning as the base of the reinforcement learning algorithm, apply the sorting problem as a case study, and assess the performance from two aspects, the time expense and the brain capacity. We compare the total number of training steps between nontransfer and transfer methods to study the efficiencies and evaluate their differences in brain capacity (i.e., the percentage of the updated Q-values in the Q-table). According to our experimental results, the difference in the total number of training steps will become smaller when the size of the numbers to be sorted increases. Our results also show that the brain capacities of transfer and nontransfer reinforcement learning will be similar when they both reach a similar training level.

In the era of Industry 4.0, manufacturing sites are becoming more sophisticated and connected wit... more In the era of Industry 4.0, manufacturing sites are becoming more sophisticated and connected with the aid of Information Technology (IT) and Industrial Internet of-ings (IIoT) infrastructure.What makes a smart manufacturing enterprise as opposed to a traditional one is the ability to solve existing problems and predict issues to fix them before they occur while creating advantaged value. Dashboards, assisting decision-makers in business markets, can organize data from machines, sensors, and workers into a real-time visual representation. -ey provide a quick overview of how your entire business is operating and where you stand in relation to the key performance indicators for supporting the development toward more resource-efficient and sustainable processes. However, if users need more information related to the indicators shown on the dashboard, current approaches mainly rely on human effort to search and retrieve the information.-e lack of interaction between the dashboard and us...

Data Driven Innovation (DDI) is one of the main pillars leading to improved productivity and econ... more Data Driven Innovation (DDI) is one of the main pillars leading to improved productivity and economic growth for companies. The adoption and use of DDI may determine their competitive advantages and innovation capabilities to exploit data. However, it is known that the key to DDI lies in the data itself. Comparing the use of data and the creation of innovation, the way how data flows, is collected and integrated is no less pressing. In this paper, we present the Data Concept Map to take into account the human factor for digitalization challenges and benefit organizations, especially for small- and medium sized enterprises, in terms of saving costs, improving flexibility and marching towards Data Driven Enterprise. We introduce the concept of utilizing smart technologies as a channel to work with humans for expediting data collection, system integration and information flow. An experimental case study is conducted for a manufacturing application to demonstrate the effectiveness of th...

Arabian Journal for Science and Engineering

This paper presents a real-life scheduling problem of minimizing total weighted tardiness on iden... more This paper presents a real-life scheduling problem of minimizing total weighted tardiness on identical parallel machines with eligibility constraints which is originated from the manufacturing plant of an industrial metal nuts company. Because the problem is NP-hard, a new electromagnetism-like mechanism algorithm is proposed to solve the problem. In the proposed algorithm, the particle is redesigned to represent jobs with valid assignment to machines. A distance measure between particles is proposed by the concept of a number-guessing game. Then, the new attraction and repulsion operators are developed to move a particle to the new particle. The computational results show that the proposed algorithm performs better than the current scheduling method of the metal nut plant and other existing algorithms.

Applied Sciences

Clinical Named Entity Recognition (CNER) focuses on locating named entities in electronic medical... more Clinical Named Entity Recognition (CNER) focuses on locating named entities in electronic medical records (EMRs) and the obtained results play an important role in the development of intelligent biomedical systems. In addition to the research in alphabetic languages, the study of non-alphabetic languages has attracted considerable attention as well. In this paper, a neural model is proposed to address the extraction of entities from EMRs written in Chinese. To avoid erroneous noise being caused by the Chinese word segmentation, we employ the character embeddings as the only feature without extra resources. In our model, concatenated n-gram character embeddings are used to represent the context semantics. The self-attention mechanism is then applied to model long-range dependencies of embeddings. The concatenation of the new representations obtained by the attention module is taken as the input to bidirectional long short-term memory (BiLSTM), followed by a conditional random field (...

Applied Sciences

Railway wheelsets are the key to ensuring the safe operation of trains. To achieve zero-defect pr... more Railway wheelsets are the key to ensuring the safe operation of trains. To achieve zero-defect production, railway equipment manufacturers must strictly control every link in the wheelset production process. The press-fit curve output by the wheelset assembly machine is an essential indicator of the wheelset’s assembly quality. The operators will still need to manually and individually recheck press-fit curves in our practical case. However, there are many uncertainties in the manual inspection. For example, subjective judgment can easily cause inconsistent judgment results between different inspectors, or the probability of human misinterpretation can increase as the working hours increase. Therefore, this study proposes an intelligent railway wheelset inspection system based on deep learning, which improves the reliability and efficiency of manual inspection of wheelset assembly quality. To solve the severe imbalance in the number of collected images, this study establishes a pred...

IEEE Access

Electrical discharge machining (EDM) can effectively solve the shortcomings of traditional machin... more Electrical discharge machining (EDM) can effectively solve the shortcomings of traditional machining processes that cannot process special materials, so it is widely used on workpieces with strong hardness materials, such as titanium alloys and tool steels to produce various molds and dies. However, the operating procedures of EDM are quite complicated and low machining productivity. To improve machining efficiency, this study develops an intelligent system that adaptively controls debris removal operations instead of using preset debris removal parameters. A feature extraction method is proposed in this study to effectively identify the machining states from streaming images of the machining curve for evaluating the appropriate time of the debris removal operation. Then, the extracted features feed into the artificial neural network model to establish a debris removal predicted model. The preliminary experimental result shows that the established predicted model can achieve an accuracy of 96.93% for a testing dataset containing 750 machining curve images. To further verify the effectiveness of the proposed intelligent system in improving EDM efficiency, we integrate the debris removal predicted model into the EDM machine and test it on the manufacturing site. Compared with the preset debris removal parameter, the proposed intelligent system can save nearly 38.60% of machining time for the machining depth of 6.45mm under specific EDM conditions.

The International Journal of Advanced Manufacturing Technology

Mobile Information Systems

In the era of Industry 4.0, manufacturing sites are becoming more sophisticated and connected wit... more In the era of Industry 4.0, manufacturing sites are becoming more sophisticated and connected with the aid of Information Technology (IT) and Industrial Internet of Things (IIoT) infrastructure. What makes a smart manufacturing enterprise as opposed to a traditional one is the ability to solve existing problems and predict issues to fix them before they occur while creating advantaged value. Dashboards, assisting decision-makers in business markets, can organize data from machines, sensors, and workers into a real-time visual representation. They provide a quick overview of how your entire business is operating and where you stand in relation to the key performance indicators for supporting the development toward more resource-efficient and sustainable processes. However, if users need more information related to the indicators shown on the dashboard, current approaches mainly rely on human effort to search and retrieve the information. The lack of interaction between the dashboard ...

Mobile Information Systems

There is a vast need for the use of digital display instruments in the manufacturing industry due... more There is a vast need for the use of digital display instruments in the manufacturing industry due to the simple operation and high precision. In addition to the numerical data acquisition, it is usually necessary to input additional text for the condition assessment as well. However, since most of these measure instruments do not provide any interfaces for users to access the values and it often lacks proper devices to input the text during the working process, these two tasks are highly human intensive under current conditions. In order to facilitate the smooth running of the work for operators, we propose a lightweight application which can be installed on smartphones or wearable devices using multidigit recognition and speech recognition techniques without changing too much of their workflow. The experimental results demonstrate that our approach can achieve high accuracy. Thus, the proposed solution can effectively resolve data input issues in the manufacturing sites, thereby re...

IEEE Access

Immediate monitoring of the conditions of the grinding wheel during the grinding process is impor... more Immediate monitoring of the conditions of the grinding wheel during the grinding process is important because it directly affects the surface accuracy of the workpiece. Because the variation in machining sound during the grinding process is very important for the field operator to judge whether the grinding wheel is worn or not, this study applies artificial intelligence technology to attempt to learn the experiences of auditory recognition of experienced operators. Therefore, we propose an intelligent system based on machining sound and deep learning to recognize the grinding wheel condition. This study uses a microphone embedded in the grinding machine to collect audio signals during the grinding process, and extracts the most discriminated feature from spectrum analysis. The features will be input the designed CNNs architecture to create a training model based on deep learning for distinguishing different conditions of the grinding wheel. Experimental results show that the proposed system can achieve an accuracy of 97.44%, a precision of 98.26% and a recall of 96.59% from 820 testing samples. INDEX TERMS Grinding wheel wear, intelligent system, machining sound, audio signals, deep learning. I. INTRODUCTION CHENG-HSIUNG LEE received the B.I.M. and M.I.M. degrees in information management from the Chaoyang University of Technology, in 2002 and 2004, respectively, and the Ph.D. degree in computer science and engineering from the

This paper presents an automated system for grading pathological images of prostatic carcinoma ba... more This paper presents an automated system for grading pathological images of prostatic carcinoma based on a set of texture features extracted by multi-categories of methods including multi-wavelets, Gabor-filters, GLCM, and fractal dimensions. We apply 5-fold cross-validation procedure to a set of 205 pathological prostate images for training and testing. Experimental results show that the fractal dimension (FD) feature set can achieve 92.7% of CCR without feature selection and 94.1% of CCR with feature selection by using support vector machine classifier. If features of multi-categories are considered and optimized, the CCR can be promoted to 95.6%. The CCR drops to 92.7% if FD-based features are removed from the combined feature set. Such a result suggests that features of FD category have significant contributions and should be included for consideration if features are selected from multi-categories.

2013 International Conference on System Science and Engineering (ICSSE), 2013

ABSTRACT In this paper, we present a classification system for differentiating malignant pulmonar... more ABSTRACT In this paper, we present a classification system for differentiating malignant pulmonary nodules from benign nodules in computed tomography (CT) images based on a set of fractal features derived from the fractional Brownian motion (fBm) model. In a set of 107 CT images obtained from 107 different patients with each image containing a solitary pulmonary nodule, our experimental result show that the accuracy rate of classification and the area under the Receiver Operating Characteristic (ROC) curve are 83.11% and 0.8437, respectively, by using the proposed fractal-based feature set and a support vector machine classifier. Such a result demonstrates that our classification system has highly satisfactory diagnostic performance by analyzing the fractal features of lung nodules in CT images taken from a single post-contrast CT scan.

This paper presents a new method to automatically grade pathological prostate images according to... more This paper presents a new method to automatically grade pathological prostate images according to Gleason grading system. Two feature extraction methods were proposed based on fractal dimension to analyze the variations of intensity and texture complexity in images. Each image can be classified into appropriate grade by using Bayes classifier and k-Nearest-Neighbor (k-NN) classifier, respectively. Leaving-One-Out approach was used to estimate the correct classification rates. Experimental results showed that 92.86 % of accuracy can be achieved by using Bayes classifier and 89.01 % of accuracy can be achieved by using k-NN classifier for a set of 182 pathological prostate images.

Computational Intelligence and Neuroscience, 2020

Previous studies have shown that training a reinforcement model for the sorting problem takes ver... more Previous studies have shown that training a reinforcement model for the sorting problem takes very long time, even for small sets of data. To study whether transfer learning could improve the training process of reinforcement learning, we employ Q-learning as the base of the reinforcement learning algorithm, apply the sorting problem as a case study, and assess the performance from two aspects, the time expense and the brain capacity. We compare the total number of training steps between nontransfer and transfer methods to study the efficiencies and evaluate their differences in brain capacity (i.e., the percentage of the updated Q-values in the Q-table). According to our experimental results, the difference in the total number of training steps will become smaller when the size of the numbers to be sorted increases. Our results also show that the brain capacities of transfer and nontransfer reinforcement learning will be similar when they both reach a similar training level.

In the era of Industry 4.0, manufacturing sites are becoming more sophisticated and connected wit... more In the era of Industry 4.0, manufacturing sites are becoming more sophisticated and connected with the aid of Information Technology (IT) and Industrial Internet of-ings (IIoT) infrastructure.What makes a smart manufacturing enterprise as opposed to a traditional one is the ability to solve existing problems and predict issues to fix them before they occur while creating advantaged value. Dashboards, assisting decision-makers in business markets, can organize data from machines, sensors, and workers into a real-time visual representation. -ey provide a quick overview of how your entire business is operating and where you stand in relation to the key performance indicators for supporting the development toward more resource-efficient and sustainable processes. However, if users need more information related to the indicators shown on the dashboard, current approaches mainly rely on human effort to search and retrieve the information.-e lack of interaction between the dashboard and us...

Data Driven Innovation (DDI) is one of the main pillars leading to improved productivity and econ... more Data Driven Innovation (DDI) is one of the main pillars leading to improved productivity and economic growth for companies. The adoption and use of DDI may determine their competitive advantages and innovation capabilities to exploit data. However, it is known that the key to DDI lies in the data itself. Comparing the use of data and the creation of innovation, the way how data flows, is collected and integrated is no less pressing. In this paper, we present the Data Concept Map to take into account the human factor for digitalization challenges and benefit organizations, especially for small- and medium sized enterprises, in terms of saving costs, improving flexibility and marching towards Data Driven Enterprise. We introduce the concept of utilizing smart technologies as a channel to work with humans for expediting data collection, system integration and information flow. An experimental case study is conducted for a manufacturing application to demonstrate the effectiveness of th...

Arabian Journal for Science and Engineering

This paper presents a real-life scheduling problem of minimizing total weighted tardiness on iden... more This paper presents a real-life scheduling problem of minimizing total weighted tardiness on identical parallel machines with eligibility constraints which is originated from the manufacturing plant of an industrial metal nuts company. Because the problem is NP-hard, a new electromagnetism-like mechanism algorithm is proposed to solve the problem. In the proposed algorithm, the particle is redesigned to represent jobs with valid assignment to machines. A distance measure between particles is proposed by the concept of a number-guessing game. Then, the new attraction and repulsion operators are developed to move a particle to the new particle. The computational results show that the proposed algorithm performs better than the current scheduling method of the metal nut plant and other existing algorithms.

Applied Sciences

Clinical Named Entity Recognition (CNER) focuses on locating named entities in electronic medical... more Clinical Named Entity Recognition (CNER) focuses on locating named entities in electronic medical records (EMRs) and the obtained results play an important role in the development of intelligent biomedical systems. In addition to the research in alphabetic languages, the study of non-alphabetic languages has attracted considerable attention as well. In this paper, a neural model is proposed to address the extraction of entities from EMRs written in Chinese. To avoid erroneous noise being caused by the Chinese word segmentation, we employ the character embeddings as the only feature without extra resources. In our model, concatenated n-gram character embeddings are used to represent the context semantics. The self-attention mechanism is then applied to model long-range dependencies of embeddings. The concatenation of the new representations obtained by the attention module is taken as the input to bidirectional long short-term memory (BiLSTM), followed by a conditional random field (...

Applied Sciences

Railway wheelsets are the key to ensuring the safe operation of trains. To achieve zero-defect pr... more Railway wheelsets are the key to ensuring the safe operation of trains. To achieve zero-defect production, railway equipment manufacturers must strictly control every link in the wheelset production process. The press-fit curve output by the wheelset assembly machine is an essential indicator of the wheelset’s assembly quality. The operators will still need to manually and individually recheck press-fit curves in our practical case. However, there are many uncertainties in the manual inspection. For example, subjective judgment can easily cause inconsistent judgment results between different inspectors, or the probability of human misinterpretation can increase as the working hours increase. Therefore, this study proposes an intelligent railway wheelset inspection system based on deep learning, which improves the reliability and efficiency of manual inspection of wheelset assembly quality. To solve the severe imbalance in the number of collected images, this study establishes a pred...

IEEE Access

Electrical discharge machining (EDM) can effectively solve the shortcomings of traditional machin... more Electrical discharge machining (EDM) can effectively solve the shortcomings of traditional machining processes that cannot process special materials, so it is widely used on workpieces with strong hardness materials, such as titanium alloys and tool steels to produce various molds and dies. However, the operating procedures of EDM are quite complicated and low machining productivity. To improve machining efficiency, this study develops an intelligent system that adaptively controls debris removal operations instead of using preset debris removal parameters. A feature extraction method is proposed in this study to effectively identify the machining states from streaming images of the machining curve for evaluating the appropriate time of the debris removal operation. Then, the extracted features feed into the artificial neural network model to establish a debris removal predicted model. The preliminary experimental result shows that the established predicted model can achieve an accuracy of 96.93% for a testing dataset containing 750 machining curve images. To further verify the effectiveness of the proposed intelligent system in improving EDM efficiency, we integrate the debris removal predicted model into the EDM machine and test it on the manufacturing site. Compared with the preset debris removal parameter, the proposed intelligent system can save nearly 38.60% of machining time for the machining depth of 6.45mm under specific EDM conditions.

The International Journal of Advanced Manufacturing Technology

Mobile Information Systems

In the era of Industry 4.0, manufacturing sites are becoming more sophisticated and connected wit... more In the era of Industry 4.0, manufacturing sites are becoming more sophisticated and connected with the aid of Information Technology (IT) and Industrial Internet of Things (IIoT) infrastructure. What makes a smart manufacturing enterprise as opposed to a traditional one is the ability to solve existing problems and predict issues to fix them before they occur while creating advantaged value. Dashboards, assisting decision-makers in business markets, can organize data from machines, sensors, and workers into a real-time visual representation. They provide a quick overview of how your entire business is operating and where you stand in relation to the key performance indicators for supporting the development toward more resource-efficient and sustainable processes. However, if users need more information related to the indicators shown on the dashboard, current approaches mainly rely on human effort to search and retrieve the information. The lack of interaction between the dashboard ...

Mobile Information Systems

There is a vast need for the use of digital display instruments in the manufacturing industry due... more There is a vast need for the use of digital display instruments in the manufacturing industry due to the simple operation and high precision. In addition to the numerical data acquisition, it is usually necessary to input additional text for the condition assessment as well. However, since most of these measure instruments do not provide any interfaces for users to access the values and it often lacks proper devices to input the text during the working process, these two tasks are highly human intensive under current conditions. In order to facilitate the smooth running of the work for operators, we propose a lightweight application which can be installed on smartphones or wearable devices using multidigit recognition and speech recognition techniques without changing too much of their workflow. The experimental results demonstrate that our approach can achieve high accuracy. Thus, the proposed solution can effectively resolve data input issues in the manufacturing sites, thereby re...

IEEE Access

Immediate monitoring of the conditions of the grinding wheel during the grinding process is impor... more Immediate monitoring of the conditions of the grinding wheel during the grinding process is important because it directly affects the surface accuracy of the workpiece. Because the variation in machining sound during the grinding process is very important for the field operator to judge whether the grinding wheel is worn or not, this study applies artificial intelligence technology to attempt to learn the experiences of auditory recognition of experienced operators. Therefore, we propose an intelligent system based on machining sound and deep learning to recognize the grinding wheel condition. This study uses a microphone embedded in the grinding machine to collect audio signals during the grinding process, and extracts the most discriminated feature from spectrum analysis. The features will be input the designed CNNs architecture to create a training model based on deep learning for distinguishing different conditions of the grinding wheel. Experimental results show that the proposed system can achieve an accuracy of 97.44%, a precision of 98.26% and a recall of 96.59% from 820 testing samples. INDEX TERMS Grinding wheel wear, intelligent system, machining sound, audio signals, deep learning. I. INTRODUCTION CHENG-HSIUNG LEE received the B.I.M. and M.I.M. degrees in information management from the Chaoyang University of Technology, in 2002 and 2004, respectively, and the Ph.D. degree in computer science and engineering from the

This paper presents an automated system for grading pathological images of prostatic carcinoma ba... more This paper presents an automated system for grading pathological images of prostatic carcinoma based on a set of texture features extracted by multi-categories of methods including multi-wavelets, Gabor-filters, GLCM, and fractal dimensions. We apply 5-fold cross-validation procedure to a set of 205 pathological prostate images for training and testing. Experimental results show that the fractal dimension (FD) feature set can achieve 92.7% of CCR without feature selection and 94.1% of CCR with feature selection by using support vector machine classifier. If features of multi-categories are considered and optimized, the CCR can be promoted to 95.6%. The CCR drops to 92.7% if FD-based features are removed from the combined feature set. Such a result suggests that features of FD category have significant contributions and should be included for consideration if features are selected from multi-categories.

2013 International Conference on System Science and Engineering (ICSSE), 2013

ABSTRACT In this paper, we present a classification system for differentiating malignant pulmonar... more ABSTRACT In this paper, we present a classification system for differentiating malignant pulmonary nodules from benign nodules in computed tomography (CT) images based on a set of fractal features derived from the fractional Brownian motion (fBm) model. In a set of 107 CT images obtained from 107 different patients with each image containing a solitary pulmonary nodule, our experimental result show that the accuracy rate of classification and the area under the Receiver Operating Characteristic (ROC) curve are 83.11% and 0.8437, respectively, by using the proposed fractal-based feature set and a support vector machine classifier. Such a result demonstrates that our classification system has highly satisfactory diagnostic performance by analyzing the fractal features of lung nodules in CT images taken from a single post-contrast CT scan.