• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.15 no.12
• /
• pp.4439-4455
• /
• 2021

Compared with two-stage object detection algorithms, one-stage algorithms provide a better trade-off between real-time performance and accuracy. However, these methods treat the intermediate features equally, which lacks the flexibility to emphasize meaningful information for classification and location. Besides, they ignore the interaction of contextual information from different scales, which is important for medium and small objects detection. To tackle these problems, we propose an image pyramid network based on dual attention mechanism (DAIPNet), which builds an image pyramid to enrich the spatial information while emphasizing multi-scale informative features based on dual attention mechanisms for one-stage object detection. Our framework utilizes a pre-trained backbone as standard detection network, where the designed image pyramid network (IPN) is used as auxiliary network to provide complementary information. Here, the dual attention mechanism is composed of the adaptive feature fusion module (AFFM) and the progressive attention fusion module (PAFM). AFFM is designed to automatically pay attention to the feature maps with different importance from the backbone and auxiliary network, while PAFM is utilized to adaptively learn the channel attentive information in the context transfer process. Furthermore, in the IPN, we build an image pyramid to extract scale-wise features from downsampled images of different scales, where the features are further fused at different states to enrich scale-wise information and learn more comprehensive feature representations. Experimental results are shown on MS COCO dataset. Our proposed detector with a 300 × 300 input achieves superior performance of 32.6% mAP on the MS COCO test-dev compared with state-of-the-art methods.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.24 no.10
• /
• pp.1262-1270
• /
• 2020

Recently, the smart video surveillance research, which has been receiving increasing attention, has mainly focused on the intruder detection and tracking, and abandoned object detection. On the other hand, research on real-time detection of stolen objects is relatively insufficient compared to its importance. Considering various smart surveillance video application environments, this paper presents two different types of stolen object detection algorithms. We first propose an algorithm that detects theft of statically and dynamically registered surveillance objects using a dual background subtraction model. In addition, we propose another algorithm that detects theft of general surveillance objects by applying the dual background subtraction model and Mask R-CNN-based object segmentation technology. The former algorithm can provide economical theft detection service for pre-registered surveillance objects in low computational power environments, and the latter algorithm can be applied to the theft detection of a wider range of general surveillance objects in environments capable of providing sufficient computational power.

• Phonetics and Speech Sciences
• /
• v.2 no.3
• /
• pp.141-148
• /
• 2010

In this paper, voice activity detection (VAD) for dual-channel noisy speech recognition is proposed in which spatial cues are employed. In the proposed method, a probability model for speech presence/absence is constructed using spatial cues obtained from dual-channel input signal, and a speech activity interval is detected through this probability model. In particular, spatial cues are composed of interaural time differences and interaural level differences of dual-channel speech signals, and the probability model for speech presence/absence is based on a Gaussian kernel density. In order to evaluate the performance of the proposed VAD method, speech recognition is performed for speech segments that only include speech intervals detected by the proposed VAD method. The performance of the proposed method is compared with those of several methods such as an SNR-based method, a direction of arrival (DOA) based method, and a phase vector based method. It is shown from the speech recognition experiments that the proposed method outperforms conventional methods by providing relative word error rates reductions of 11.68%, 41.92%, and 10.15% compared with SNR-based, DOA-based, and phase vector based method, respectively.

• Journal of Institute of Control, Robotics and Systems
• /
• v.4 no.2
• /
• pp.264-272
• /
• 1998

This paper proposes a fault detection and recovery mechanism for a fault-tolerant Mini-MAP system, and provides detailed techniques for its implementation. This paper considers the fault-tolerant Mini-MAP system which has dual layer structure from the LLC sublayer down to the physical layer to cope with the faults of those layers. For a good fault detection, a redundant and hierarchical fault supervision architecture is proposed and its implementation technique for a stable detection operation is provided. Information for the fault location is provided from data reported with a fault detection and obtained by an additional network diagnosis. The faults are recovered by the stand-by sparing method applied for a dual network composed of two equivalent networks. A network switch mechanism is proposed to achieve a reliable and stable network function. A fault-tolerant Mini-MAP system is implemented by applying the proposed fault detection and recovery mechanism.

• Journal of the Korean Society for Precision Engineering
• /
• v.25 no.8
• /
• pp.29-36
• /
• 2008

An optical rotary encoder is easy to implement for automation system applications. In particular, the output of the encoder has a digital form pulse, which is also easy to be connected to a popular digital controller. By using an incremental encoder and a counting device, it is easy to measure angular displacement, as the number of the output pulses is proportional to the rotational displacement. This method can only detect the angular placement once a pulse signal comes out of the encoder. The angular displacement detection period is strongly subject to the change of the angular displacement in case of ultimate low velocity range. They have ultimate long detection period or cannot even detect angular displacement at near zero velocity. This paper proposes an algorithm for detecting angular displacement by using a dual encoder system with two encoders of normal resolution. The angular displacement detecting algorithm is able to keep detection period moderately at near zero velocity and even detect constant angular displacement within nominal period. It is useful for motion control applications in case of changing rotational direction at which there occurs zero velocity. In this paper, various experimental results are shown for the angular displacement detection algorithm.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.62 no.1
• /
• pp.120-126
• /
• 2013

This paper proposes a novel scheme of fault detection and self-recovery for space-borne dual ring counters subject to transient faults. The considered ring counter is equipped with hardware redundancy, but it has a limited output domain where direct access to the current state is unavailable. We employ the theory of corrective control to detect any transient fault occurring to the counter bits and to realize immediate self-recovery of the ring counter back to the normal state. The structure of the fault-tolerant controller is designed to be minimal regardless of the counter size. To validate the applicability, we implement the proposed system on a commercial FGPA board.

• 제어로봇시스템학회:학술대회논문집
• /
• 1990.10a
• /
• pp.773-778
• /
• 1990

In this paper, a reliable control system structured with dual CPU modules and dual I/O modules is implemented as a means of achieving a highly reliable fault tolerant control system. For this, faults in the system modules are first examined, and a fault detection technique consisting of self diagnostic, comparison process, and exception processing is applied. Also reliability analysis is conducted for the discrete time Markov model with dual structure. It is shown quantitatively that the reliability is improved in the control system with dual structure in comparison with a system with single module structure.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.29 no.12C
• /
• pp.1660-1668
• /
• 2004

This paper proposes a Voice Activity Detection (VAD) algorithm based on Radial Basis Function (RBF) network using dual threshold. The k-means clustering and Least Mean Square (LMS) algorithm are used to upade the RBF network to the underlying speech condition. The inputs for RBF are the three parameters in a Code Exited Linear Prediction (CELP) coder, which works stably under various background noise levels. Dual hangover threshold applies in BRF-VAD for reducing error, because threshold value has trade off effect in VAD decision. The experimental result show that the proposed VAD algorithm achieves better performance than G.729 Annex B at any noise level.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.34 no.7C
• /
• pp.715-723
• /
• 2009

In this paper, we propose a new detection method called dual-cell combining (DCC) detection for the acquisition in time of spread spectrum codes in the presence of residual frequency offset (RFO). When the RFO exists, the correlation peak used for detection during the acquisition process is split into two neighboring peaks with smaller amplitudes, resulting in considerable degradation in the overall acquisition performance of conventional methods. In the DCC detection method, the decision variable for detection is formed by combining two consecutive correlator outputs so that the influence of the reduction in the correlation peak due to the RFO can be alleviated. Numerical results show that the DCC detection method can offer better mean-time-to-synchrouization performance than the conventional method based on the cell-by-cell detection.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.54 no.1
• /
• pp.121-128
• /
• 2017

In this paper, we propose a dual-channel acoustic event detection (AED) method using nonnegative tensor factorization (NTF) and hidden Markov model (HMM) in order to improve detection accuracy of AED in multisource environments. The proposed method first detects multiple acoustic events by utilizing channel gains obtained from the NTF technique applied to dual-channel input signals. After that, an HMM-based likelihood ratio test is carried out to verify the detected events by using channel gains. The detection accuracy of the proposed method is measured by F-measures under 9 different multisource conditions. Then, it is also compared with those of conventional AED methods such as Gaussian mixture model and nonnegative matrix factorization. It is shown from the experiments that the proposed method outperforms the convectional methods under all the multisource conditions.