• Journal of Web Engineering
• /
• v.21 no.2
• /
• pp.497-522
• /
• 2021

Audio event classification refers to the detection and classification of non-verbal signals, such as dog and horn sounds included in audio data, by a computer. Recently, deep neural network technology has been applied to audio event classification, exhibiting higher performance when compared to existing models. Among them, a convolutional neural network (CNN)-based training method that receives audio in the form of a spectrogram, which is a two-dimensional image, has been widely used. However, audio event classification has poor performance on test data when it is recorded by a device (unknown device) different from that used to record training data (known device). This is because the frequency range emphasized is different for each device used during recording, and the shapes of the resulting spectrograms generated by known devices and those generated by unknown devices differ. In this study, to improve the performance of the event classification system, a CNN based on the log mel-spectrogram separation technique was applied to the event classification system, and the performance of unknown devices was evaluated. The system can classify 16 types of audio signals. It receives audio data at 0.4-s length, and measures the accuracy of test data generated from unknown devices with a model trained via training data generated from known devices. The experiment showed that the performance compared to the baseline exhibited a relative improvement of up to 37.33%, from 63.63% to 73.33% based on Google Pixel, and from 47.42% to 65.12% based on the LG V50.

• Speech Sciences
• /
• v.15 no.2
• /
• pp.55-65
• /
• 2008

Recently, the multi-model based speech recognizer has been used quite successfully for noisy speech recognition. For the selection of the reference HMM (hidden Markov model) which best matches the noise type and SNR (signal to noise ratio) of the input testing speech, the estimation of the SNR value using the VAD (voice activity detection) algorithm and the classification of the noise type based on the GMM (Gaussian mixture model) have been done separately in the multi-model framework. As the SNR estimation process is vulnerable to errors, we propose an efficient method which can classify simultaneously the SNR values and noise types. The KL (Kullback-Leibler) distance between the single Gaussian distributions for the noise signal during the training and testing is utilized for the classification. The recognition experiments have been done on the Aurora 2 database showing the usefulness of the model compensation method in the multi-model based speech recognizer. We could also see that further performance improvement was achievable by combining the probability density function of the MCT (multi-condition training) with that of the reference HMM compensated by the D-JA (data-driven Jacobian adaptation) in the multi-model based speech recognizer.

• KIPS Transactions on Software and Data Engineering
• /
• v.4 no.7
• /
• pp.283-290
• /
• 2015

In this paper, we propose a hybrid method for the recognition of answer types in the WiseQA system. The answer types are classified into two categories: the lexical answer type (LAT) and the semantic answer type (SAT). This paper proposes two models for the LAT detection. One is a rule-based model using question focuses. The other is a machine learning model based on sequence labeling. We also propose two models for the SAT classification. They are a machine learning model based on multiclass classification and a filtering-rule model based on the lexical answer type. The performance of the LAT detection and the SAT classification shows F1-score of 82.47% and precision of 77.13%, respectively. Compared with IBM Watson for the performance of the LAT, the precision is 1.0% lower and the recall is 7.4% higher.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.15 no.6
• /
• pp.2168-2187
• /
• 2021

The smart grid replaces the traditional power structure with information inventiveness that contributes to a new physical structure. In such a field, malicious information injection can potentially lead to extreme results. Incorrect, FDI attacks will never be identified by typical residual techniques for false data identification. Most of the work on the detection of FDI attacks is based on the linearized power system model DC and does not detect attacks from the AC model. Also, the overwhelming majority of current FDIA recognition approaches focus on FDIA, whilst significant injection location data cannot be achieved. Building on the continuous developments in deep learning, we propose a Deep Learning based Locational Detection technique to continuously recognize the specific areas of FDIA. In the development area solver gap happiness is a False Data Detector (FDD) that incorporates a Convolutional Neural Network (CNN). The FDD is established enough to catch the fake information. As a multi-label classifier, the following CNN is utilized to evaluate the irregularity and cooccurrence dependency of power flow calculations due to the possible attacks. There are no earlier statistical assumptions in the architecture proposed, as they are "model-free." It is also "cost-accommodating" since it does not alter the current FDD framework and it is only several microseconds on a household computer during the identification procedure. We have shown that ANN-MLP, SVM-RBF, and CNN can conduct locational detection under different noise and attack circumstances through broad experience in IEEE 14, 30, 57, and 118 bus systems. Moreover, the multi-name classification method used successfully improves the precision of the present identification.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.26 no.2
• /
• pp.313-316
• /
• 2022

In this paper, we implement a system that detects abnormalities in the charging data transmitted from the charger during the charging process of electric vehicles and controls them remotely. Using classification algorithms such as logistic regression, KNN, SVM, and decision trees, to do this, an analysis model is created that judges the data received from the charger as normal and abnormal. In addition, a model is created to determine the cause of the abnormality using the existing charging data based on the analysis of the type of charger abnormality. Finally, it is solved using unsupervised learning method to find new patterns of abnormal data.

• The Journal of the Convergence on Culture Technology
• /
• v.10 no.4
• /
• pp.575-580
• /
• 2024

Detecting the status of websites, normal or phishing, is necessary to defend against intelligent phishing attacks. We propose a machine learning-based classification to predict the status of websites. First, we collect information about 'URL', convert it into numerical data, and remove outliers. Second, we apply VIF(Variance Inflation Factors) to understand the correlation and independence between variables. Finally, we develop a phishing website detection model with machine learning-based classifications, which predicts website status. In the test datasets, Random Forest showed the best performance, with precision of 93.74%, recall of 92.26%, and accuracy of 93.14%. In the future, we expect to apply our model to detect various phishing crimes.

• Proceedings of the KSRS Conference
• /
• 2008.10a
• /
• pp.410-413
• /
• 2008

Existing Outlier detections have been widely studied in geosensor networks. Recently, machine learning and data mining have been applied the outlier detection method to build a model that distinguishes outliers based on anchored criterion. However, it is difficult for the existing methods to detect outliers against incoming time-variant data, because outlier detection needs to monitor incoming data and classify irregular attacks. Therefore, in order to solve the problem, we propose a time-variant outlier detection using 2-dimensional grid method based on unanchored criterion. In the paper, outliers using geosensor data was performed to classify efficiently. The proposed method can be utilized applications such as network intrusion detection, stock market analysis, and error data detection in bank account.

• Smart Structures and Systems
• /
• v.13 no.5
• /
• pp.849-868
• /
• 2014

An indirect approach is explored for structural health bridge monitoring allowing for wide, yet cost-effective, bridge stock coverage. The detection capability of the approach is tested in a laboratory setting for three different reversible proxy types of damage scenarios: changes in the support conditions (rotational restraint), additional damping, and an added mass at the midspan. A set of frequency features is used in conjunction with a support vector machine classifier on data measured from a passing vehicle at the wheel and suspension levels, and directly from the bridge structure for comparison. For each type of damage, four levels of severity were explored. The results show that for each damage type, the classification accuracy based on data measured from the passing vehicle is, on average, as good as or better than the classification accuracy based on data measured from the bridge. Classification accuracy showed a steady trend for low (1-1.75 m/s) and high vehicle speeds (2-2.75 m/s), with a decrease of about 7% for the latter. These results show promise towards a highly mobile structural health bridge monitoring system for wide and cost-effective bridge stock coverage.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.17 no.5
• /
• pp.1433-1449
• /
• 2023

In recent years, the rapid development of social networks has led to a rapid increase in the amount of information available on the Internet, which contains a large amount of sensitive information related to pornography, politics, and terrorism. In the aspect of sensitive image detection, the existing machine learning algorithms are confronted with problems such as large model size, long training time, and slow detection speed when auditing and supervising. In order to detect sensitive images more accurately and quickly, this paper proposes a multiclassification sensitive image detection method based on lightweight Convolutional Neural Network. On the basis of the EfficientNet model, this method combines the Ghost Module idea of the GhostNet model and adds the SE channel attention mechanism in the Ghost Module for feature extraction training. The experimental results on the sensitive image data set constructed in this paper show that the accuracy of the proposed method in sensitive information detection is 94.46% higher than that of the similar methods. Then, the model is pruned through an ablation experiment, and the activation function is replaced by Hard-Swish, which reduces the parameters of the original model by 54.67%. Under the condition of ensuring accuracy, the detection time of a single image is reduced from 8.88ms to 6.37ms. The results of the experiment demonstrate that the method put forward has successfully enhanced the precision of identifying multi-class sensitive images, significantly decreased the number of parameters in the model, and achieved higher accuracy than comparable algorithms while using a more lightweight model design.

• Journal of the Korea Institute of Military Science and Technology
• /
• v.25 no.6
• /
• pp.572-579
• /
• 2022

The analysis of foreign aircraft appearing suddenly in air defense identification zones requires a lot of cost and time. This study aims to develop a pre-trained model that can identify neighboring military aircraft based on aircraft photographs available on the web and present a model that can determine which aircraft corresponds to based on aerial photographs taken by allies. The advantages of this model are to reduce the cost and time required for model classification by proposing a pre-trained model and to improve the performance of the classifier by data augmentation of edge-detected images, cropping, flipping and so on.