• Journal of Digital Convergence
• /
• v.11 no.6
• /
• pp.169-174
• /
• 2013

Speech recognition, the problem of performance degradation is the difference between the model training and recognition environments. Silence features normalized using the method as a way to reduce the inconsistency of such an environment. Silence features normalized way of existing in the low signal-to-noise ratio. Increase the energy level of the silence interval for voice and non-voice classification accuracy due to the falling. There is a problem in the recognition performance is degraded. This paper proposed a robust speech detection method in noisy environments using a silence feature normalization and voice energy maximize. In the high signal-to-noise ratio for the proposed method was used to maximize the characteristics receive less characterized the effects of noise by the voice energy. Cepstral feature distribution of voice / non-voice characteristics in the low signal-to-noise ratio and improves the recognition performance. Result of the recognition experiment, recognition performance improved compared to the conventional method.

• Journal of the Korea Convergence Society
• /
• v.8 no.5
• /
• pp.13-17
• /
• 2017

Existing Speech feature extracting method in speech Signal, there are incorrect recognition rates due to incorrect speech which is not clear threshold value. In this article, the modeling method for improving speech recognition performance that combines the feature extraction for speech and silence characteristics normalized to the non-speech. The proposed method is minimized the noise affect, and speech recognition model are convergence of speech signal feature extraction to each speech frame and the silence feature normalization. Also, this method create the original speech signal with energy spectrum similar to entropy, therefore speech noise effects are to receive less of the noise. the performance values are improved in signal to noise ration by the silence feature normalization. We fixed speech and non speech classification standard value in cepstrum For th Performance analysis of the method presented in this paper is showed by comparing the results with CHMM HMM, the recognition rate was improved 2.7%p in the speech dependent and advanced 0.7%p in the speech independent.

• Journal of Sensor Science and Technology
• /
• v.20 no.2
• /
• pp.137-144
• /
• 2011

Wireless sensor network(WSN) has the potential to greatly effect many aspects of u-healthcare. By outfitting the potential with WSN, wearable sensor node can collects real-time data on physiological status and transmits through base station to server PC. However, there is a significant gap between WSN and healthcare. WSN has the limited resource about computing capability and data transmission according to bio-sensor sampling rates and channels to apply healthcare system. If a wearable node transmits ECG and accelerometer data of 4 channel sampled at 100 Hz, these data may occur high loss packets for transmitting human activity and ECG to server PC. Therefore current wearable sensor nodes have to solve above mentioned problems to be suited for u-healthcare system. Most WSN based activity and ECG monitoring system have been implemented some algorithms which are applied for signal vector magnitude(SVM) algorithm and ECG noise algorithm in server PC. In this paper, A wearable sensor node using integrated ECG and 3-axial accelerometer based on wireless sensor network is designed and developed. It can form multi-hop network with relay nodes to extend network range in WSN. Our wearable nodes can transmit 1-channel activity data processed activity classification data vector using SVM algorithm to 3-channel accelerometer data. ECG signals are contaminated with high frequency noise such as power line interference and muscle artifact. Our wearable sensor nodes can remove high frequency noise to clear original ECG signal for healthcare monitoring.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.24 no.6
• /
• pp.700-705
• /
• 2020

Generative adversarial networks are methods of generating images by opposing two neural networks. When generating the image, randomly generated noise is rearranged to generate the image. The image generated by this method is not generated well depending on the noise, and it is difficult to generate a proper image when the number of pixels of the image is small In addition, the speed and size of data accumulation in data classification increases, and there are many difficulties in labeling them. In this paper, to solve this problem, we propose a technique to generate noise based on random noise using real data. Since the proposed system generates an image based on the existing image, it is confirmed that it is possible to generate a more natural image, and if it is used for learning, it shows a higher hit rate than the existing method using the hostile neural network respectively.

• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.15 no.1
• /
• pp.27-35
• /
• 2022

In this paper we propose white noise adding method to prevent missclassification of deep learning system by adversarial attacks. The proposed method is that adding white noise to input image that is benign or adversarial example. The experimental results are showing that the proposed method is robustness to 3 adversarial attacks such as FGSM attack, BIN attack and CW attack. The recognition accuracies of Resnet model with 18, 34, 50 and 101 layers are enhanced when white noise is added to test data set while it does not affect to classification of benign test dataset. The proposed model is applicable to defense to adversarial attacks and replace to time- consuming and high expensive defense method against adversarial attacks such as adversarial training method and deep learning replacing method.

• Journal of the Korean Society of Marine Environment & Safety
• /
• v.27 no.7
• /
• pp.1088-1097
• /
• 2021

Vibration data of mechanical equipment inevitably have noise. This noise adversely af ects the maintenance of mechanical equipment. Accordingly, the performance of a learning model depends on how effectively the noise of the data is removed. In this study, the noise of the data was removed using the Denoising Auto Encoder (DAE) technique which does not include the characteristic extraction process in preprocessing time series data. In addition, the performance was compared with that of the Wavelet Transform, which is widely used for machine signal processing. The performance comparison was conducted by calculating the failure detection rate. For a more accurate comparison, a classification performance evaluation criterion, the F-1 Score, was calculated. Failure data were detected using the One-Class SVM technique. The performance comparison, revealed that the DAE technique performed better than the Wavelet Transform technique in terms of failure diagnosis and error rate.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.22 no.2
• /
• pp.233-242
• /
• 2018

Previous works for detecting arrhythmia have mostly used nonlinear method such as artificial neural network, fuzzy theory, support vector machine to increase classification accuracy. Most methods require higher computational cost and larger processing time. Therefore it is necessary to design efficient algorithm that classifies PVC(premature ventricular contraction) and decreases computational cost by accurately detecting feature point based on only R peak through optimal R wave. For this purpose, we detected R wave through optimal threshold value and extracted RR interval and R peak pattern from noise-free ECG signal through the preprocessing method. Also, we classified PVC in realtime through RR interval and R peak pattern. The performance of R wave detection and PVC classification is evaluated by using 9 record of MIT-BIH arrhythmia database that included over 30. The achieved scores indicate the average of 99.02% in R wave detection and the rate of 94.85% in PVC classification.

• The KIPS Transactions:PartB
• /
• v.16B no.5
• /
• pp.377-384
• /
• 2009

A new algorithm in order to classify various contents in the image documents, such as text, figure, graph, table, etc. is proposed in this paper by classifying contents using texture-based PCA, and by segmenting document images using local entropy-based histogram. Local entropy and histogram made the binarization of image document not only robust to various transformation and noise, but also easy and less time-consuming. And texture-based PCA algorithm for each segmented region was taken notice of each content in the image documents having different texture information. Through this, it was not necessary to establish any pre-defined structural information, and advantages were found from the fact of fast and efficient classification. The result demonstrated that the proposed method had shown better performances of segmentation and classification for various images, and is also found superior to previous methods by its efficiency.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.38 no.4
• /
• pp.442-452
• /
• 2001

In this paper, we proposed a postprocessing algorithm for quantization effects reduction in block coded images using the block classification and adaptive filtering. The proposed method consists of classification, adaptive inter-block filtering, and intra-block filtering. First, each block is classified into one of seven classes based on the characteristics of 8$\times$8 DCT coefficients. Then each block boundary is filtered by adaptive inter-block fitters according to the block classification. finally for blocks which are classified into edge block, intra-block filtering is performed. Experimental results show that the proposed method gives better results than the conventional methods from both a subjective and an objective viewpoint.

• The Journal of the Acoustical Society of Korea
• /
• v.39 no.5
• /
• pp.406-413
• /
• 2020

In this paper, to improve the classification accuracy of bird and amphibian acoustic sound, we utilize GLU (Gated Linear Unit) and Self-attention that encourages the network to extract important features from data and discriminate relevant important frames from all the input sequences for further performance improvement. To utilize acoustic data, we convert 1-D acoustic data to a log-Mel spectrogram. Subsequently, undesirable component such as background noise in the log-Mel spectrogram is reduced by GLU. Then, we employ the proposed temporal self-attention to improve classification accuracy. The data consist of 6-species of birds, 8-species of amphibians including endangered species in the natural environment. As a result, our proposed method is shown to achieve an accuracy of 91 % with bird data and 93 % with amphibian data. Overall, an improvement of about 6 % ~ 7 % accuracy in performance is achieved compared to the existing algorithms.