scispace - formally typeset
Search or ask a question
Institution

Mitsubishi

CompanyTokyo, Japan
About: Mitsubishi is a company organization based out in Tokyo, Japan. It is known for research contribution in the topics: Layer (electronics) & Signal. The organization has 53115 authors who have published 54821 publications receiving 870150 citations. The organization is also known as: Mitsubishi Group of Companies & Mitsubishi Companies.


Papers
More filters
Proceedings ArticleDOI
04 May 2014
TL;DR: Full recurrent connections are added to certain hidden layer of a conventional feedforward DNN and allow the model to capture the temporal dependency in deep representations to achieve state-of-the-art performance without front-end preprocessing, speaker adaptive training or multiple decoding passes.
Abstract: In this work, we propose recurrent deep neural networks (DNNs) for robust automatic speech recognition (ASR). Full recurrent connections are added to certain hidden layer of a conventional feedforward DNN and allow the model to capture the temporal dependency in deep representations. A new backpropagation through time (BPTT) algorithm is introduced to make the minibatch stochastic gradient descent (SGD) on the proposed recurrent DNNs more efficient and effective. We evaluate the proposed recurrent DNN architecture under the hybrid setup on both the 2 nd CHiME challenge (track 2) and Aurora-4 tasks. Experimental results on the CHiME challenge data show that the proposed system can obtain consistent 7% relative WER improvements over the DNN systems, achieving state-of-the-art performance without front-end preprocessing, speaker adaptive training or multiple decoding passes. For the experiments on Aurora-4, the proposed system achieves 4% relative WER improvement over a strong DNN baseline system.

140 citations

Journal ArticleDOI
16 Feb 1994
TL;DR: In this paper, a 64-kb DRAM with a boost-level generator with body contact structure and reduced body-effect of sense-amplifier transistors is presented.
Abstract: For future ULSI DRAMs beyond the 256 Mb generation, several circuit techniques and memory cell structures have been proposed to meet the requirement of high performance at low voltage. These solutions frequently involve complicated processing steps and/or the ultimate limitations of current Si-MOS devices. DRAM on silicon on insulator (SOI) substrate is a more simple solution to the problem. Thin-film SOI structures with isolation by implanted oxygen (SIMOX) process are under investigation for SRAM and logic. A SOI-DRAM test device with 100 nm thick SOI film has been fabricated in 0.5 /spl mu/m CMOS/SIMOX technology. With this 64 kb SOI-DRAM the bit-line to memory cell capacitance ratio Cb/Cs is reduced by 25% compared with the reference bulk-Si DRAM, because of the decreased junction capacitance. RAS access time tRAC is 70 ns at 2.7 VVcc, as fast as the equivalent bulk-Si device at 4 VVcc. The clock timing in this DRAM is not optimized, so access time should improve with well-tuned clocks. The boosted-level generator with body-contact structure enhances the upper Vcc margin and the reduced body-effect of sense-amplifier transistors improves the lower Vcc margin. The SOI-DRAM has an operating Vcc range from 2.3 V to 4.0 V. >

140 citations

Journal ArticleDOI
TL;DR: This paper proposes a generalized probability density function based on the nth power of a cosine-squared function that derives the average covariance matrix for various different elementary scatterers and shows that the result has a very simple analytical form suitable for use in model-based decomposition schemes.
Abstract: Current polarimetric model-based decomposition techniques are limited to specific types of vegetation because of their assumptions about the volume scattering component. In this paper, we propose a generalized probability density function based on the nth power of a cosine-squared function. This distribution is completely characterized by two parameters; a mean orientation angle and the power of the cosine-squared function. We show that the underlying randomness of the distribution is only a function of the power of the cosine-squared function. We then derive the average covariance matrix for various different elementary scatterers showing that the result has a very simple analytical form suitable for use in model-based decomposition schemes.

140 citations

Patent
07 Apr 1995
TL;DR: In this paper, a gloss-providing layer consisting of a pigment and a synthetic polymer latex as a binder and having a glossy surface with a 75° specular gloss of at least 25% was proposed.
Abstract: An ink jet recording sheet comprising a support, at least one ink-receiving layer formed on the support, and a gloss-providing layer formed on the ink-receiving layer, said ink-receiving layer consisting essentially of a pigment and a binder, and said gloss-providing layer consisting essentially of a pigment and a synthetic polymer latex as a binder and having a glossy surface with a 75° specular gloss of at least 25% as stipulated in JIS-Z8741, wherein at least 70 parts by weight in 100 parts by weight of the pigment in the gloss-providing layer are constituted by colloidal particles having an average particle size of at most 300 nm.

140 citations

Patent
28 Sep 1989
TL;DR: A shape memory polymer foam which takes on a deformed shape and an as-molded shape was shown to recover its original shape when the polymer foam is heated to a temperature higher than the glass transition point (Tg) of the polymer as discussed by the authors.
Abstract: A shape memory polymer foam which takes on a deformed shape and an as-molded shaped, said deformed shape being produced when the polymer foam is compressed at a temperature higher than the glass transition point (Tg) of the polymer and then kept compressed at a temperature lower than the Tg until compression sets, said as-molded shape being produced when the compressed polymer foam is heated again to a temperature higher than the Tg until it recovers its original shape.

140 citations


Authors

Showing all 53117 results

NameH-indexPapersCitations
Thomas S. Huang1461299101564
Kazunari Domen13090877964
Kozo Kaibuchi12949360461
Yoshimi Takai12268061478
William T. Freeman11343269007
Tadayuki Takahashi11293257501
Takashi Saito112104152937
H. Vincent Poor109211667723
Qi Tian96103041010
Andreas F. Molisch9677747530
Takeshi Sakurai9549243221
Akira Kikuchi9341228893
Markus Gross9158832881
Eiichi Nakamura9084531632
Michael Wooldridge8754350675
Network Information
Related Institutions (5)
Osaka University
185.6K papers, 5.1M citations

90% related

Tokyo Institute of Technology
101.6K papers, 2.3M citations

89% related

University of Tokyo
337.5K papers, 10.1M citations

89% related

Nagoya University
128.2K papers, 3.2M citations

88% related

Kyoto University
217.2K papers, 6.5M citations

88% related

Performance
Metrics
No. of papers from the Institution in previous years
YearPapers
20231
20222
2021199
2020310
2019389
2018422