• ETRI Journal
• /
• v.45 no.2
• /
• pp.318-328
• /
• 2023

Recently, embedded systems, such as mobile platforms, have multiple processing units that can operate in parallel, such as centralized processing units (CPUs) and neural processing units (NPUs). We can use deep-learning compilers to generate machine code optimized for these embedded systems from a deep neural network (DNN). However, the deep-learning compilers proposed so far generate codes that sequentially execute DNN operators on a single processing unit or parallel codes for graphic processing units (GPUs). In this study, we propose PartitionTuner, an operator scheduler for deep-learning compilers that supports multiple heterogeneous PUs including CPUs and NPUs. PartitionTuner can generate an operator-scheduling plan that uses all available PUs simultaneously to minimize overall DNN inference time. Operator scheduling is based on the analysis of DNN architecture and the performance profiles of individual and group operators measured on heterogeneous processing units. By the experiments for seven DNNs, PartitionTuner generates scheduling plans that perform 5.03% better than a static type-based operator-scheduling technique for SqueezeNet. In addition, PartitionTuner outperforms recent profiling-based operator-scheduling techniques for ResNet50, ResNet18, and SqueezeNet by 7.18%, 5.36%, and 2.73%, respectively.

• Smart Media Journal
• /
• v.5 no.3
• /
• pp.35-41
• /
• 2016

In this paper, we propose hand-gesture signal recognition based on EPS(Electronic Potential Sensor) using Deep learning model. Extracted signals which from Electronic field based sensor, EPS have much of the noise, so it must remove in pre-processing. After the noise are removed with filter using frequency feature, the signals are reconstructed with dimensional transformation to overcome limit which have just one-dimension feature with voltage value for using convolution operation. Then, the reconstructed signal data is finally classified and recognized using multiple learning layers model based on deep learning. Since the statistical model based on probability is sensitive to initial parameters, the result can change after training in modeling phase. Deep learning model can overcome this problem because of several layers in training phase. In experiment, we used two different deep learning structures, Convolutional neural networks and Recurrent Neural Network and compared with statistical model algorithm with four kinds of gestures. The recognition result of method using convolutional neural network is better than other algorithms in EPS gesture signal recognition.

• Journal of Korea Multimedia Society
• /
• v.23 no.8
• /
• pp.927-939
• /
• 2020

Gait recognition can identify people's identity from a long distance, which is very important for improving the intelligence of the monitoring system. Among many human features, gait features have the advantages of being remotely available, robust, and secure. Traditional gait feature extraction, affected by the development of behavior recognition, can only rely on manual feature extraction, which cannot meet the needs of fine gait recognition. The emergence of deep convolutional neural networks has made researchers get rid of complex feature design engineering, and can automatically learn available features through data, which has been widely used. In this paper,conduct feature metric learning in the three-dimensional space by combining the three-dimensional convolution features of the gait sequence and the Siamese structure. This method can capture the information of spatial dimension and time dimension from the continuous periodic gait sequence, and further improve the accuracy and practicability of gait recognition.

• ETRI Journal
• /
• v.40 no.5
• /
• pp.643-652
• /
• 2018

Dysarthria is a degenerative disorder of the central nervous system that affects the control of articulation and pitch; therefore, it affects the uniqueness of sound produced by the speaker. Hence, dysarthric speaker recognition is a challenging task. In this paper, a feature-extraction method based on deep belief networks is presented for the task of identifying a speaker suffering from dysarthria. The effectiveness of the proposed method is demonstrated and compared with well-known Mel-frequency cepstral coefficient features. For classification purposes, the use of a multi-layer perceptron neural network is proposed with two structures. Our evaluations using the universal access speech database produced promising results and outperformed other baseline methods. In addition, speaker identification under both text-dependent and text-independent conditions are explored. The highest accuracy achieved using the proposed system is 97.3%.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.14 no.6
• /
• pp.2444-2463
• /
• 2020

Significant detection of video can more rationally allocate computing resources and reduce the amount of computation to improve accuracy. Deep learning can extract the edge features of the image, providing technical support for video saliency. This paper proposes a new detection method. We combine the Convolutional Neural Network (CNN) and the Deep Bidirectional LSTM Network (DB-LSTM) to learn the spatio-temporal features by exploring the object motion information and object motion information to generate video. A continuous frame of significant images. We also analyzed the sample database and found that human attention and significant conversion are time-dependent, so we also considered the significance detection of video cross-frame. Finally, experiments show that our method is superior to other advanced methods.

• The Korean Journal of Applied Statistics
• /
• v.33 no.5
• /
• pp.615-625
• /
• 2020

The document-term frequency matrix is a term extracted from documents in which the group information exists in text mining. In this study, we generated the document-term frequency matrix for document classification according to research field. We applied the traditional term weighting function term frequency-inverse document frequency (TF-IDF) to the generated document-term frequency matrix. In addition, we applied term frequency-inverse gravity moment (TF-IGM). We also generated a document-keyword weighted matrix by extracting keywords to improve the document classification accuracy. Based on the keywords matrix extracted, we classify documents using a deep neural network. In order to find the optimal model in the deep neural network, the accuracy of document classification was verified by changing the number of hidden layers and hidden nodes. Consequently, the model with eight hidden layers showed the highest accuracy and all TF-IGM document classification accuracy (according to parameter changes) were higher than TF-IDF. In addition, the deep neural network was confirmed to have better accuracy than the support vector machine. Therefore, we propose a method to apply TF-IGM and a deep neural network in the document classification.

• Journal of Auto-vehicle Safety Association
• /
• v.13 no.4
• /
• pp.33-38
• /
• 2021

This paper presents a comparative study between the parameter-optimized Pacejka model and artificial neural network model for the tire force estimation. The two different approaches are investigated and compared in this study. First, offline optimization is conducted based on Pacejka Magic Formula model to determine the proper parameter set for the minimization of tire force error between the model and test data set. Second, deep neural network model is used to fit the model to the tire test data set. The actual tire forces are measured using MTS Flat-Track test platform and the measurements are used as the reference tire data set. The focus of this study is on the applicability of machine learning technique to tire force estimation. It is shown via the regression results that the deep neural network model is more effective in describing the tire force than the parameter-optimized Pacejka model.

• The Journal of the Acoustical Society of Korea
• /
• v.42 no.4
• /
• pp.357-363
• /
• 2023

In underwater signal processing, separating individual signals from mixed signals has long been a challenge due to low signal quality. The common method using Short-time Fourier transform for spectrogram analysis has faced criticism for its complex parameter optimization and loss of phase data. We propose a Triple-path Recurrent Neural Network, based on the Dual-path Recurrent Neural Network's success in long time series signal processing, to handle three-dimensional tensors from multi-channel sensor input signals. By dividing input signals into short chunks and creating a 3D tensor, the method accounts for relationships within and between chunks and channels, enabling local and global feature learning. The proposed technique demonstrates improved Root Mean Square Error and Scale Invariant Signal to Noise Ratio compared to the existing method.

• International journal of advanced smart convergence
• /
• v.8 no.2
• /
• pp.147-154
• /
• 2019

In this paper, we conducted a study that utilizes deep learning to calculate appropriate physical exercise information when basic human factors such as sex, age, height, and weight of users come in. To apply deep learning, a method was applied to calculate the amount of fat needed to calculate the amount of one repetition maximum by utilizing the structure of the basic Deep Neural Network. By applying Accuracy improvement methods such as Relu, Weight initialization, and Dropout to existing deep learning structures, we have improved Accuracy to derive a lean body weight that is closer to actual results. In addition, the results were derived by applying a formula for calculating the one repetition maximum load on upper and lower body movements for use in actual physical exercise. If studies continue, such as the way they are applied in this paper, they will be able to suggest effective physical exercise options for different conditions as well as conditions for users.

• International journal of advanced smart convergence
• /
• v.10 no.3
• /
• pp.51-58
• /
• 2021

In this paper, we propose a deep learning classifier for estimating the number of layers in the Earth's structure. When installing a grounding system, knowledge of the subsurface in the area is absolutely necessary. The subsurface structure can be modeled by the earth parameters. Knowing the exact number of layers can significantly reduce the amount of computation to estimate these parameters. The classifier consists of a feedforward neural network. Apparent resistivity curves were used to train the deep learning classifier. The apparent resistivity at 20 equally spaced log points in each curve are used as the features for the input of the deep learning classifier. Apparent resistivity curve data sets are collected either by theoretical calculations or by Wenner's measurement method. Deep learning classifiers are coded by Keras, an open source neural network library written in Python. This model has been shown to converge with close to 100% accuracy.