• IEMEK Journal of Embedded Systems and Applications
• /
• v.17 no.6
• /
• pp.367-374
• /
• 2022

The acceleration of neural networks has become an important topic in the field of computer vision. An accelerator is absolutely necessary for accelerating the lightweight model. Most accelerator-supported operators focused on direct convolution operations. If the accelerator does not provide GEMM operation, it is mostly replaced by CPU operation. In this paper, we proposed an optimization technique for 2-stage tiling-based GEMM routines on VTA. We improved performance of the matrix multiplication routine by maximizing the reusability of the input matrix and optimizing the operation pipelining. In addition, we applied the proposed technique to the DarkNet framework to check the performance improvement of the matrix multiplication routine. The proposed GEMM method showed a performance improvement of more than 2.4 times compared to the non-optimized GEMM method. The inference performance of our DarkNet framework has also improved by at least 2.3 times.

• Smart Media Journal
• /
• v.13 no.6
• /
• pp.16-23
• /
• 2024

Pigs express their reactions to their environment and health status through a variety of sounds, such as grunting, coughing, and screaming. Given the significance of pig vocalizations, their study has recently become a vital source of data for livestock industry workers. To facilitate this, we propose a lightweight deep learning model based on MobileNet that analyzes pig vocal patterns to distinguish pig voices from farm noise and differentiate between vocal sounds and coughing. This model was able to accurately identify pig vocalizations amidst a variety of background noises and cough sounds within the pigsty. Test results demonstrated that this model achieved a high accuracy of 98.2%. Based on these results, future research is expected to address issues such as analyzing pig emotions and identifying stress levels.

• Journal of IKEEE
• /
• v.28 no.2
• /
• pp.168-173
• /
• 2024

Speaker recognition refers to a technology that analyzes voice frequencies that are different for each individual and compares them with pre-stored voices to determine the identity of the person. Deep learning-based speaker recognition is being applied to many fields, and pet robots are one of them. However, the hardware performance of pet robots is very limited in terms of the large memory space and calculations of deep learning technology. This is an important problem that pet robots must solve in real-time interaction with users. Lightening deep learning models has become an important way to solve the above problems, and a lot of research is being done recently. In this paper, we describe the results of research on lightweight speaker recognition for pet robots by constructing a voice data set for pet robots, which is a specific command type, and comparing the results of models using residuals. In the conclusion, we present the results of the proposed method and Future research plans are described.

• PNF and Movement
• /
• v.22 no.1
• /
• pp.43-54
• /
• 2024

Purpose: Falls among persons older than 65 years are a significant concern due to their frequency and severity. This study aimed to develop a vest-type embedded artificial intelligence (AI) system capable of detecting and predicting falls in various scenarios. Methods: In this study, we established and developed a vest-type embedded AI system to judge and predict falls in various directions and situations. To train the AI, we collected data using acceleration and gyroscope values from a six-axis sensor attached to the seventh cervical and the second sacral vertebrae of the user, considering accurate motion analysis of the human body. The model was constructed using a neural network-based AI prediction algorithm to anticipate the direction of falls using the collected pedestrian data. Results: We focused on developing a lightweight and efficient fall prediction model for integration into an embedded AI algorithm system, ensuring real-time network optimization. Our results showed that the accuracy of fall occurrence and direction prediction using the trained fall prediction model was 89.0% and 78.8%, respectively. Furthermore, the fall occurrence and direction prediction accuracy of the model quantized for embedded porting was 87.0 % and 75.5 %, respectively. Conclusion: The developed fall detection and prediction system, designed as a vest-type with an embedded AI algorithm, offers the potential to provide real-time feedback to pedestrians in clinical settings and proactively prepare for accidents.

• Annual Conference of KIPS
• /
• 2019.05a
• /
• pp.435-437
• /
• 2019

모바일 환경에서 다양한 AI 관련 응용을 수행하기 위해, 정확도에 기반한 크고 깊은 신경망 이외에, 정확도를 비교적 유지하면서 좀더 효율적인 신경망 구조에 대한 다양한 연구가 진행중이다. 본 논문에서는 모바일 딥러닝을 위한 다양한 임베디드 장치 및 모바일 폰에서의 성능 평가를 통해 경량 신경망의 비교 분석에 대한 연구를 담고 있다.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.26 no.5
• /
• pp.668-674
• /
• 2022

Machine learning techniques using visual data have high usability in fields of industry and service such as scene recognition, fault detection, security and user analysis. Among these, user analysis through the videos from CCTV is one of the practical way of using vision data. Also, many studies about lightweight artificial neural network have been published to increase high usability for mobile and embedded environment so far. In this study, we propose the network combining the object detection and classification for mobile graphic processing unit. This network detects pedestrian and face, classifies age and gender from detected face. Proposed network is constructed based on MobileNet, YOLOv2 and skip connection. Both detection and classification models are trained individually and combined as 2-stage structure. Also, attention mechanism is used to improve detection and classification ability. Nvidia Jetson Nano is used to run and evaluate the proposed system.

• Korean Journal of Remote Sensing
• /
• v.38 no.2
• /
• pp.167-177
• /
• 2022

In order to obtain satellite image products using the image transmitted to the ground station after capturing the satellite images, many image pre/post-processing steps are involved. During the pre/post-processing, when converting from level 1R images to level 1G images, geometric correction is essential. An interpolation method necessary for geometric correction is inevitably used, and the quality of the level 1G images is determined according to the accuracy of the interpolation method. Also, it is crucial to speed up the interpolation algorithm by the level processor. In this paper, we proposed a lightweight CNN-based interpolation method required for geometric correction when converting from level 1R to level 1G. The proposed method doubles the resolution of satellite images and constructs a deep learning network with a lightweight deep convolutional neural network for fast processing speed. In addition, a feature map fusion method capable of improving the image quality of multispectral (MS) bands using panchromatic (PAN) band information was proposed. The images obtained through the proposed interpolation method improved by about 0.4 dB for the PAN image and about 4.9 dB for the MS image in the quantitative peak signal-to-noise ratio (PSNR) index compared to the existing deep learning-based interpolation methods. In addition, it was confirmed that the time required to acquire an image that is twice the resolution of the 36,500×36,500 input image based on the PAN image size is improved by about 1.6 times compared to the existing deep learning-based interpolation method.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.21 no.1
• /
• pp.93-98
• /
• 2021

Due to the development of deep learning and AI, the scale of the model has grown, and it has been integrated into other fields to blend into our lives. However, in environments with limited resources such as embedded devices, it is exist difficult to apply the model and problems such as power shortages. To solve this, lightweight methods such as clouding or offloading technologies, reducing the number of parameters in the model, or optimising calculations are proposed. In this paper, quantization of learned models is applied to ONNX models used in various framework interchange formats, neural network structure and inference performance are compared with existing models, and various module methods for quantization are analyzed. Experiments show that the size of weight parameter is compressed and the inference time is more optimized than before compared to the original model.

• Transactions of the Korean Society for Noise and Vibration Engineering
• /
• v.14 no.2
• /
• pp.119-127
• /
• 2004

Butterfly valves are popularly used in service in the industrial and water works pipeline systems with large diameter because of its lightweight, simple structure and the rapidity of its manipulation. Sometimes cavitation can occur. resulting in noise, vibration and rapid deterioration of the valve trim, and do not allow further operation. Thus, the monitoring of cavitation is of economic interest and is very importance in industry. This paper proposes a condition monitoring scheme using statistical feature evaluation and support vector machine (SVM) to detect the cavitation conditions of butterfly valve which used as a flow control valve at the pumping stations. The stationary features of vibration signals are extracted from statistical moments. The SVMs are trained, and then classify normal and cavitation conditions of control valves. The SVMs with the reorganized feature vectors can distinguish the class of the untrained and untested data. The classification validity of this method is examined by various signals that are acquired from butterfly valves in the pumping stations and compared the classification success rate with those of self-organizing feature map neural network.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.32 no.11
• /
• pp.990-996
• /
• 2008

Nowadays, the trend to a lightweight design accelerates the use of advanced high strength steel (AHSS) in automotive industry. Springback phenomena is a hot issue in the sheet metal forming, especially bending process using AHSS. Several analytical methods for that have been proposed in recent years. Each of method has their advantages and disadvantages. There are only a few optimal solutions which can minimize the two objectives simultaneously. In this study, an effective method optimized the multi objective value. The method by the design of experiments(DOE) and artificial neural network(ANN) was presented to compensate springback of bending parts. This method was applied to L and V bending process. The effective method could be optimized to multiple object. It was confirmed that the proposed method was more efficient than traditional manual FEA procedure and the trial and error approach for springback compensation.