Yasuhiro Takano

Yasuhiro Takano

Auto-generated profile, based on publicly available information

Yasuhiro Takano

Chitose Institute of Science and Technology, Chitose, Japan

Research Focus & Metrics

Academic Career Length

9Years (2018-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
29 March 2018 in IEEE Signal Processing Letters
Verified
Journal Article
Closed access
YTYasuhiro Takano
Yasuhiro Takano
HSHsuan-Jung Su
Hsuan-Jung Su
Turbo channel estimation algorithms can improve estimation accuracy in multiuser (MU) multi-input multi-output (MIMO) systems. However, the conventional least squares (LS) turbo channel estimation technique based on the maximum likelihood approach is difficult to be performed in large MU-MIMO systems since it requires a cubic complexity order O(κ 3 N T 3 N R 3 ) for a system constant κ and numbers of N T -transmit and N R -receive antennas. This is because the conventional technique computes inversion of a covariance matrix whose size is proportional to κN T N R . We propose a new low-complexity LS estimation technique that requires a complexity order O(κ 3 N T 3 ) only by utilizing algebraic property on the covariance matrix. Simulation results shown in this letter verify that the proposed technique significantly reduces processing time without sacrificing the estimation performance.

11 December 2019 in IEEE Transactions on Signal Processing
Verified
Journal Article
Closed access
YTYasuhiro Takano
Yasuhiro Takano
HSHsuan-Jung Su
Hsuan-Jung Su
YSYoshiaki Shiraishi
Yoshiaki Shiraishi
MMMasakatu Morii
Masakatu Morii
Spatial-temporal (ST) subspace-based channel estimation techniques formulated with L2 minimum mean square error (MMSE) criterion can alleviate the multi-access interference (MAI) problem when the interested signals exhibit low-rank property. However, the conventional L2 ST subspace-based methods suffer from mean squared error (MSE) deterioration in unknown interference channels, due to the difficulty to separate the interested signals from the channel covariance matrices (CCMs) contaminated with unknown interference. As a solution to the problem, we propose a new L1 regularized ST channel estimation algorithm by applying the expectation-maximization (EM) algorithm to iteratively examine the signal subspace and the corresponding sparse-supports. The new algorithm updates the CCM independently of the slot-dependent L1 regularization, which enables it to correctly perform the sparse-independent component analysis (ICA) with a reasonable complexity order. Simulation results shown in this paper verify that the proposed technique significantly improves MSE performance in unknown interference MIMO channels, and hence, solves bit error rate (BER) floor problems from which the conventional receivers suffer.

21 July 2020 in IEEE Access
Verified
Journal Article
Open access
MAMazen ALOWISH
Mazen ALOWISH
YSYoshiaki Shiraishi
Yoshiaki Shiraishi
YTYasuhiro Takano
Yasuhiro Takano
MMMasami Mohri
Masami Mohri
MMMasakatu Morii
Masakatu Morii
Content retrieval is becoming an emerging application on an integrated network structure referred to as Software Defined Network and Vehicular Ad hoc Network (SDVN). The SDVN is expected to flexibly provide efficient packet routing without deploying road side units. In such the infrastructure-less environment, however, it is challenging to retrieve contents from dynamic networks clustered with vehicles on roadways. As a solution to the problem, we propose a new network design by conjoining Named Data Network (NDN) with SDVN (NDN-SDVN). The new NDN-SDVN is designed as to improve stableness of the cluster formation and hence enables to efficiently perform the content retrieval over Vehicle-to-Vehicle (V2V) communications. Note that one of the most important problems in the V2V is how to select a cluster head. The proposed NDN-SDN adaptively determines the cluster head by using the SDN controller under the maximum Quality of Service (QoS) criterion, while reserving an assistance head based on a Moth Flame Optimization (MFO) algorithm to immediately recompose a new cluster with the back-off time. Moreover, the content retrieval in the NDN-SDVN can be performed over both intra- and inter-cluster communications. To this end, a Kalman filter with Neural Network (KF-NN) is applied for selecting a certain number of preferable vehicles in 1-hop in order to mitigate the broadcast storm problem that often occurs in an NDN. Content retrieval is initiated by sending interest packets including two additional fields of the timestamp and the target MAC address. The head vehicle of the SDN controller maintains ontology regarding the content information cashed by other vehicles, which significantly improves the packet delivery ratio. This NDN-SDVN design is developed on the OMNeT++ simulator and their advantageous results are evaluated in terms of the head lifetime, the cluster lifetime, the satisfactory rate, the latency and the packet delivery ratio.

1 January 2019 in Journal of Information Processing
Verified
Journal Article
Open access
PDF file
TNTatsuya Nagai
Tatsuya Nagai
MTMakoto TAKITA
Makoto TAKITA
KFKeisuke Furumoto
Keisuke Furumoto
YSYoshiaki Shiraishi
Yoshiaki Shiraishi
KXKelin Xia
Kelin Xia
MMMasakatu Morii
Masakatu Morii
Organizations are plagued by sophisticated and diversified cyber attacks. In order to prevent such attacks, it is necessary to understand threat trends and to take measures to protect their assets. Security vendors publish reports which contain threat trends or analysis of malware. These reports are useful for help in responding to a cyber security incident. However, it is difficult to collect threat information from multiple sources such as security blog posts. In this paper, we propose a method to efficiently collect information from the relationship between words using SeededLDA. In our case studies, we visualize the relationship between the words from security blog posts which were published in 2017 by eight security vendors, and demonstrate how our method helps to understand threat trends in the IoT industry and financial institutions.

7 August 2024 in IEEE Access
Verified
Journal Article
Open access
CLChu-Tung Liu
Chu-Tung Liu
HSHsuan-Jung Su
Hsuan-Jung Su
HTHsien-Bo Tseng
Hsien-Bo Tseng
JLJhe-Yi Lin
Jhe-Yi Lin
WWWei-Hsiang Wang
Wei-Hsiang Wang
YTYasuhiro Takano
Yasuhiro Takano
Device-to-Device (D2D) communications facilitate direct data exchange between two devices, bypassing the necessity of routing through a central base station. This proximity-based communication offers advantages such as high data rates, low latency, reduced power consumption, and efficient frequency reuse. In the cellular spectrum, D2D operations encompass overlay D2D and underlay D2D. Overlay D2D allocates orthogonal radio resources to D2D and cellular users, resulting in inefficient frequency reuse. On the other hand, underlay D2D shares the same radio resources among D2D and cellular users, necessitating centralized coordination and extensive exchanges of channel and traffic information for effective interference management. To address the limitations of overlay and underlay D2D, we introduce a scheme that allows D2D users to simultaneously and non-orthogonally transmit both D2D and cellular signals. This approach resembles underlay D2D in terms of high frequency reuse but eliminates the need for centralized interference management. In this paper, we investigate various system design aspects based on this scheme, including power allocation, multiple-antenna extension, scheduling constraints, and different interference cancellation techniques. Incorporating a coordinated scheduling method, the proposed scheme enhances the sum rate of a conventional multiple-antenna overlay D2D system by 2.14 times, with only a negligible impact on the cellular communication rate. Furthermore, the proposed scheme improves the sum rate by up to 87% within conventional underlay D2D systems, all without the need for additional centralized coordination.

15 December 2023
Verified
Conference Paper
Open access
RMReiya Muraguchi
Reiya Muraguchi
YTYasuhiro Takano
Yasuhiro Takano
HSHsuan-Jung Su
Hsuan-Jung Su
Classification-based localization approaches can improve the estimation accuracy according to resolution of fingerprinting dataset. However, the approaches may not always be applicable for wide area or for pursuing accuracy since the cost required to make a database is proportional to the resolution of received signal strength indicator (RSSI) fingerprints. As a solution to the problem, we propose a new localization technique using convolution neural network (CNN)-aided regression. Specifically, a ridge regression is performed via a CNN architecture by inputting Hankel matrix constructed from RSSI samples, while improving the over-fitting issues by utilizing the max-pooling and dropout layers. The experiment results using real measurement data verified that, even when a training dataset with a moderate resolution of grid interval = 3m is assumed, the proposed method achieves mean absolute error (MAE) of 1.7m, which is 53% of that of the conventional K-nearest neighbors (KNN) technique.