Journal of Signal Processing Vol. 21(2017) No. 4 Special Issue on Papers Awarded the Student Paper Award at NCSP'17 (Editor-in-Chief: Keikichi Hirose, Editor: Yoshikazu Miyanaga, Guest Editor: Masashi Unoki, Honorary Editor-in-Chief: Takashi Yahagi)
In this study, we investigate the synchronization phenomena in two rings of van der Pol oscillators coupled by resistors. We propose a novel coupled oscillatory system comprising two rings of van der Pol oscillators with different coupling schema. We focus on the coupling strengths of the coupled van der Pol oscillators. By computer simulation, we investigate how the synchronization phenomena change by changing the coupling strengths. In these results, we observe various synchronization phenomena.
In order to reduce the volume of test data, built-in self test (BIST) and BIST-aided scan test (BAST) techniques have been proposed. To provide the test pattern generated by an automatic test pattern generator (ATPG) using BAST, we enhanced the structure of a pseudorandom pattern generator (PRPG) by inserting MUXes and NOT gates in the linear feedback shift register (LFSR) based on correlations of ATPG patterns. The procedures can achieve about 15 to 56% reduction in the volume of test data for BAST.
In this paper, we propose an immediate cooperative line wait time estimation system using Bluetooth low energy (BLE, marketed as Bluetooth Smart) on a smartphone; this system is a modified version of our previous proposed method. To estimate the wait time, we utilize time stamps of when users approach and move past two preinstalled receivers. Our system comprises three main components: the receivers, a wait time estimator, and a database. The receivers record two types of data: the recorded time and the RSSI values. The wait time estimator uses the wait time estimation algorithm, which includes three main subroutines: the maximum RSSI decision, in-decision, and out-decision on the receivers for each user's smartphone. By calibrating and analyzing the recorded log data, the wait time estimator estimates the estimated wait time. This estimated wait time is stored in the database, then provided through the website to the queueing users. The experimental results showed that the difference between the estimated wait time and the expected wait time was within 10 s for all measurements.
Recently, services using position information have been increasing. Radio waves and ultrasonic waves are used to acquire position information. Ultrasonic waves can be transmitted and received with existing speakers and popular smartphones. Therefore, there is no need for new devices for position estimation and their convenience is high. However, to the best of our knowledge, a modulation method and a symbol synchronization method for ultrasonic communication that are applicable to existing speakers and smartphones have not been proposed. In this paper, we propose an ultrasonic communication method applicable to existing speakers and smartphones. Moreover, we propose a method of acquiring position information from the signal level of ultrasonic waves at a receiver based on the proposed ultrasonic communication method. Through an experimental evaluation, we confirmed the effectiveness of communications and position estimation.
With the diversification of mobile communication systems, we need to calculate the exact bit error rate (BER) under Rician fading for the multi-level modulation methods. The method for calculating the BER under Rician fading by approximating the probability density function (pdf) in the Rician distribution by the Nakagami-m distribution has the problem that their cumulative distribution functions (cdfs) do not match. In this paper, we propose a method for deriving the BER by polynomial expansion of the Rician distribution, then we show the effectiveness of the proposed equation from simulation results.
Recently, Content-Centric Networking (CCN) has emerged as a new networking paradigm. The main features of CCN architecture are in-network caching and content-based routing. Cache replacement and decision policies have been discussed as important mechanisms of in-network caching in past literature. However, these policies cannot efficiently support streaming. In this paper, we propose an in-network caching method for reducing playback interruption time of on-demand streaming over CCN through efficient use of network resources. The proposed method splits streaming content data into chunks with fixed size and stores these chunks into the cache storage of each CCN router in order to reduce playback interruption time for streaming. Experimental evaluations show that the proposed in-network caching method outperforms in-network caching methods of the original CCN architecture in terms of playback interruption time.
A delay/disruption-tolerant network (DTN) is a network that can operate in degraded environments, enabling end-to-end communication in discontinuous networks. A DTN uses a store-and-forward method. In this method, stored packets are conveyed by some terminals and forwarded to other ones. If there are packets having different priority, a low-priority packet may be pushed out from a buffer by a high-priority one. A crucial issue is that the arrival rate of low-priority packets is decreased by such pushing out. In this paper, we propose a method considering the number of packet replications. When the buffer is full, the node determines the packet to be discarded based on both the priority of the packet and the number of replications. By computer simulation, we show that the proposed method can increase the arrival rate of both low- and high-priority packets.
We evaluate our previously reported joint power and frequency-domain inter-cell interference coordination (ICIC) method for heterogeneous networks where low transmission-power pico base stations (BSs) overlay onto a high transmission-power macro BS. In the previously reported method, to protect users connected to a pico BS from interference from a high-power macro BS, we define a protected band, which is exclusively used by only pico BSs. In the remaining non-protected band, we employ soft fractional frequency reuse (SFR) at the macro BSs so that macro-to-macro ICIC is achieved. With this frequency usage, we employ adaptive transmission power control for the nonprotected band at the macro BS and simultaneously control the bandwidth allocation to the protected band from the viewpoint of proportional fairness (PF). The previously reported method is achieved with very limited information exchange between BSs. The transmission power control is a decentralized approach. Thus, in this method, we use a cost function in the objective function to lower the transmission power of the macro BSs so that the interference to the other BSs is considered. In our previous investigation, we used different values of the cost for different system conditions. In this paper, we use the same value of the cost for all evaluated conditions to obtain more realistic results. Computer simulation results show in detail the effectiveness of our method compared with conventional approaches under various system conditions.
Nowadays, a convolutional neural network (CNN) is considered as a deep learning method for image and voice recognition. A CNN can achieve higher recognition accuracy than other approaches since it can automatically extract features by its learning procedure. However, the training procedure of a CNN is time-consuming. Since the functions of a CNN are close to those of a human brain, when a CNN is applied to a complex application, it must be trained by a large amount of training data, resulting in the size of the CNN becoming huge. To train such a huge neural network by computers, a tremendous amount of training time is required. In this paper, an efficient approach is proposed that can markedly reduce the training time while only slightly sacrificing the recognition accuracy of the training procedure.
The spread of viruses, such as flu and SARS, can be modeled using social networks. In this study, we modeled the spread of a virus in a social network and attempted to suppress the virus by immunizing nodes. Although a suppression method for virus diffusion using only local information has been proposed, we proposed a new suppression method using degree and the importance of lines centrality. We evaluated the performance of the proposed method for scale-free networks. In addition, we investigated whether our proposed method is valid for various types of social networks. In numerical experiments, the proposed method showed better performance than the conventional suppression method.
This paper presents an alternative and efficient one-dimensional direction-of-arrival estimation method for wide-band sources. The proposed method is compatible with most classical subspace-based methods, such as, conventional and root multiple signal classification. Although it only employs a Gaussian mixture model with a maximum likelihood estimation algorithm, it is sufficient for exhibiting wide-band sources angle estimation. Modification of the microphone array data model is proposed and investigated in order to avoid confusion caused by unwanted side lobes in uniform linear arrays radiation. The performance is evaluated in terms of the root-mean-squared error over a range of the signal-to-noise ratio. In conclusion, the proposed method enables the synthesis of signal sources and provides a potential alternative to intelligent source localization systems.
It is well known that the interaural time difference, interaural level difference and spectral cues are used to determine three-dimensional sound localization in binaural hearing. In the case of monaural hearing, the interaural time difference and interaural level difference are not used. Therefore, it is assumed that there is a different perception of sound localization between binaural and monaural hearing. In this study, we investigate the difference in the horizontal localization of sound images and sources in monaural hearing. An experiment involving horizontal sound localization was performed with one female participant suffering from congenital complete hearing loss in the left ear. The experimental system consisted of 12 loudspeakers placed horizontally on the circumference of a circle having a radius of 1 m at 30° intervals. Four experimental sessions were performed (including 60 white-noise stimuli per session). Excluding the instances with no localization (12%), all sound images were localized on the right side (0-180°). It appeared that sound images were localized on the side with the normal-hearing ear but not on the side with the deaf ear. Sound source localization was possible generally over 360° (with ± 30° allowance, 90.8%). As a result, we confirmed that the localization of sound images and sources was different in congenital monaural hearing.
This study investigates acoustic variations when producing Lombard speech under the effect of a changing environment to identify adaptive tendencies of intelligibility. Analyses of the acoustic features of duration, F0, formants, spectral tilts and modulation spectrum in a dataset of speech at noise levels of - ∞, 66, 72, 78, 84, and 90 dB were carried out. The results show that the recognized tendencies (neutral-Lombard distinction), including lengthening vowel duration, increasing F0, shifting F1 and decreasing spectral tilts (A1-A3) are preserved among Lombard speech produced in backgrounds with a various noise levels. Our new findings are an abrupt change in F0 at 84 dB, increasing formant amplitudes, and H1-H2 variation, and a raised modulation spectrum. On the basis of physiological and psychological knowledge, we can give reasons for their correlations with intelligibility. Moreover, these variations continuously vary with increasing noise level. As a result, it is suggested that they are related to the adaptive tendencies of intelligibility.
This paper describes path-planning and traveling-control algorithms for a pesticide-spraying robot in a greenhouse. In order to search for a suitable path, we applied graph theory and expressed a greenhouse map as a set of nodes and branches. The robot searches for the path from the start node to the goal node through all branches that it needs to spray. Moreover, the robot can identify its position on a map by detecting the shapes of plant beds and walls using a laser range finder (LRF) and can decide which direction to turn. In addition, if its pesticide tank is empty, the robot needs to return to the charge node to obtain more pesticide, and then restarts traveling and spraying. We consider the validity of the path planning and traveling control from simulation results and experimental results obtained using a mock-up model of a greenhouse lane.
This paper describes the action control rules of each robot in a multiple mobile robot system for the panel cruising problem. To achieve a task by a multiple mobile robot system, it is important to decide the action control rules to avoid collision among the robots. We propose action control rules based on an evaluation function to decide the moving direction. To confirm the flexibility of the proposed action control rules, we focus on the evenness of panel points, which are passing counts to panels, and the energy consumption as the task efficiency. Moreover, we discuss the task efficiency for several workspaces and numbers of robots by comparison of the proposed rule with a simpler rule based on only the panel points.
This paper describes a prediction method for wind speed fluctuation using a deep belief network (DBN) trained with ensemble learning. In particular, we investigate the usefulness of the ensemble learning for an prediction accuracy improvement of wind speed fluctuation. Bootstrap aggregating (the bagging method), which is a typical algorithm of ensemble learning, has been applied to train the DBN. The prediction result is decided by a majority vote of each DBN output. In addition, two bagging methods with different selection methods of training data have been proposed. These proposed methods have been evaluated from several prediction results by comparison with a conventional method.
For a general tele-existence robot (TER), it is difficult for the operator to recognize the circumstances around the TER by observing the narrow view from the camera. To overcome this difficulty, we focused on a new method that provides two-dimensional (2D) map based on a laser range finder (LRF) in addition to the camera view. Existing simple 2D map methods have not been sufficiently considered from the view point of the operator's spatial perception; hence, in this study, a new remote manipulation system providing a real-time blurred 2D map is presented. Next, the operator's spatial perception using the 2D map was investigated through a tele-operation experiment by changing the scan area of the LRF and the accumulated time used to make motion blur on the 2D map. The conditions employed to make the blurred 2D map were changed among nine cases combining three scan ranges and three accumulated times. The spatial perception was also investigated by analyzing distances measured by the LRF. By comparing three distance data groups using the Kruskal-Wallis test, significant differences in the minimum distance were confirmed when the scan ranges were 1, 3 and 5m (p<.001). Also, significant differences in the distance were confirmed when the accumulated times were 0.1, 0.3 and 0.5s (p<.001). Therefore, it was confirmed that the operator's spatial perception was improved by increasing the scan range and the motion blur using the 2D map.
We propose real-time camera position and posture estimation with six degrees of freedom (6-DoF) based on local patches of an image sequence. Our method alternately performs camera motion calculation and depth map reconstruction. Using the reconstructed depth map on only edge regions of images, our estimation method was 30% faster than that based on all pixels and achieved robust motion tracking compared with the feature-points-based method.
In this paper, we investigate a blind watermarking algorithm based on a highpass filter for three-dimensional (3-D) meshes. For improving watermark detection in correlation-based watermarking, our scheme employs the highpass filter which emphasizes an impulse signal embedded as a signature into a host mesh. In the proposed method, we align the host mesh by the principal component analysis and convert from orthogonal coordinates to polar coordinates. After this preprocessing, we map the 3-D data onto a 2-D space via block segmentation and average operation, and rearrange for the 2-D data to an 1-D sequence. On the 1-D space, we apply a complex smear transform and a highpass filter. From the resulting signal, we derive the optimum complex-valued impulse signal in terms of the Euclidean norm. To generate a watermark with desirable properties, similar to a pseudonoise signal, we perform a complex desmear transform, which is the inverse system to the complex smear transform, on the complex-valued impulse signal. After reordering into the 2-D signal and 3-D mapping from the 2-D space, the watermark is embedded into the host mesh and the resulting mesh is converted to orthogonal coordinates. At the decoder, we implement an inverse process with the highpass filter for stego meshes and detect a position of the maximum value as a signature. For a 3-D Bunny model, detection rates are shown to evaluate the performance of the proposed algorithm.
In this paper, we present a novel method for estimating levels of group involvement. With the recent development of single-person analysis in computer vision, social group analysis has received growing attention and many group detection methods have been proposed. Most of the previous studies considered each person in an image in terms of binary levels of involvement (group member or not), but actually each person can have various statuses in a social space. These complexity of social status sometimes causes a decrease in the group detection accuracy. Our approach expresses each person in terms of social involvement features representing the relationship to the surrounding people. An involvement level classifier is trained by using a machine learning algorithm. We evaluated our proposed method by comparison with a previous method and confirmed the advantageousness of our method.
In this paper, a robust similar image retrieval method using extracted object features is proposed. A local texture feature as well as global contour and color features are extracted from an object image to generate unified robust feature vectors. To show the effectiveness of our method, experimental noisy image retrieval was executed using standard object image databases. The recall rate, precision rate and F-measure obtained by cross-validation were calculated to evaluate the performance of object image retrieval. High-performance image retrieval was achieved compared with the conventional methods without using combined robust features of extracted objects.
In the past, many studies have been carried out on eye-gaze input; however, in this study, we developed an eye-glance input interface that tracks a combination of short eye movements. Unlike eye-gaze input that requires high accuracy measurements, eye-glance input can be detected with only a rough indication of the direction of the eye movements, making it possible to operate even terminals with small screens, such as smartphones. In this study, we used an inexpensive camera to measure eye movements and analyzed its output using the OpenCV, an open source computer vision and machine learning software library, to construct an inexpensive and non-contact interface. In a previous study, we developed an algorithm that detected eye-glance input through image analysis using OpenCV, and fed the result of the algorithm back to our subjects. In that study, the average detection rate for the eye-glance input was 76 %. However, we also observed several problems with the algorithm, particularly the problem of false detections due to blinking of the eyes, and implemented solutions for improvement. In this study, we have made improvement with respect to the unsatisfactory detection rate recorded in our previous study, and addressed problems related to user convenience.
July 14, 2017 Due to the maintenance‚following linking services will not be available on Jul 27 from 10:00 to 15:00 (JST)(Jul 27‚ from 1:00 to 6:00(UTC)). We apologize for the inconvenience. a)reference linking b)cited-by linking c)linking to J-STAGE with JOI/OpenURL
July 03, 2017 There had been a service stop from Jul 2, 2017, 8:06 to Jul 2, 2017, 19:12(JST) (Jul 1, 2017, 23:06 to Jul 2, 2017, 10:12(UTC)) . The service has been back to normal.We apologize for any inconvenience this may cause you.
May 18, 2016 We have released “J-STAGE BETA site”.
May 01, 2015 Please note the "spoofing mail" that pretends to be J-STAGE.