Shitomi T.,NHK Japan Broadcasting Corporation Science and Technology Research Laboratories |
Murayama K.,NHK Science and Technical Research Laboratories |
Taguchi M.,NHK Science and Technical Research Laboratories |
Asakura S.,NHK Science and Technical Research Laboratories |
Shibuya K.,NHK Science and Technical Research Laboratories
IEEE International Symposium on Broadband Multimedia Systems and Broadcasting, BMSB | Year: 2012
We have developed a transmission system using ultra-multilevel orthogonal frequency division multiplexing (OFDM) technology, dual-polarized multiple-input multiple-output (MIMO) technology, and Low-Density Parity-Check (LDPC) Codes for experimental purposes on the basis of the conventional digital terrestrial broadcasting transmission scheme, ISDB-T. We conducted two field experiments: one at 23 reception points around NHK Science & Technology Research Laboratories (STRL) using an experimental transmitter installed at NHK STRL, and the other on the lawn at NHK STRL using circularly or skew polarized waves in addition to a conventional polarization set, horizontal and vertical. As a result, we obtained the required field strength in an urban environment when using the 4096QAM carrier modulation scheme with a LDPC code. Furthermore, we confirmed the advantage of circularly or skew polarized waves in an environment in which the received power differed between the horizontally and vertically polarized waves. © 2012 IEEE.
Imai A.,NHK Japan Broadcasting Corporation Science and Technology Research Laboratories |
Imai A.,Tokyo University of Science |
Tazawa N.,NHK Engineering Services Inc. |
Takagi T.,NHK Engineering Services Inc. |
And 2 more authors.
IEEE Transactions on Consumer Electronics | Year: 2013
In this paper, we describe an adaptive speech rate control technology for ultrafast listening that is equivalent to skimming. Nowadays, listening to audio books on mobile devices is quite common. Therefore, in the future we will obtain ever-increasing amounts of information through speech instead of conventional printed materials. People read books at various levels of detail from close reading to skimming. Although a similar feature to skimming is required to efficiently obtain information from audio sources, there is no tool equivalent to skimming for audio playback. Thus, we have developed a new speech rate conversion method to efficiently obtain information from audio sources with very fast replay. This algorithm will help not only sighted people to enjoy audio books but also visually impaired people because almost all of their information is obtained from speech. Thus, the implementation of this technology on special audio players for visually impaired people as a new replay function is expected to be useful. Moreover, this technology should be useful for all audio book listeners, not only people with limited sight. 1 © 1975-2011 IEEE.
Tazawa N.,NHK Engineering Services Inc. |
Totihara S.,Keio Research Institute at SFC |
Iwahana Y.,NHK Engineering Services Inc. |
Imai A.,NHK Engineering Services Inc. |
And 2 more authors.
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) | Year: 2010
We have performed studies and evaluation experiments of more acceptable rapid speech, aimed at implementation in applications such as installation in commercial devices designed for visually impaired persons. Although our method's playback time is same as the conventional High-speed playback technology, listener might feel playing in slower speeds and listen words clearer. Our proposal technology makes it possible to adaptive speech rate control in utterance position and pitch/power in the speech information, instead of changing the speed of the utterance uniformly. We performed an experiment in Japanese and American subject with visual impairments respectively to compare the conventional high-speed playback technology and that of our adaptive high-speed playback technology in terms of "Listenability". The reaction to our proposal method from subjects with visual impairments has been very positive underscoring its potential as an effective tool for listening to high-speed speech. © 2010 Springer-Verlag Berlin Heidelberg.
Segi H.,NHK Japan Broadcasting Corporation Science and Technology Research Laboratories |
Onoe K.,NHK Japan Broadcasting Corporation Science and Technology Research Laboratories |
Sato S.,NHK Japan Broadcasting Corporation Science and Technology Research Laboratories |
Kobayashi A.,NHK Japan Broadcasting Corporation Science and Technology Research Laboratories |
Ando A.,University of Toyoma
Journal of Information Technology Research | Year: 2014
Tied-mixture HMMs have been proposed as the acoustic model for large-vocabulary continuous speech recognition and have yielded promising results. They share base-distribution and provide more flexibility in choosing the degree of tying than state-clustered HMMs. However, it is unclear which acoustic models to superior to the other under the same training data. Moreover, LBG algorithm and EM algorithm, which are the usual training methods for HMMs, have not been compared. Therefore in this paper, the recognition performance of the respective HMMs and the respective training methods are compared under the same condition. It was found that the number of parameters and the word error rate for both HMMs are equivalent when the number of codebooks is sufficiently large. It was also found that training method using the LBG algorithm achieves a 90% reduction in training time compared to training method using the EM algorithm, without degradation of recognition accuracy. Copyright © 2014, IGI Global.
Park H.,NHK Japan Broadcasting Corporation Science and Technology Research Laboratories |
Mitsumine H.,NHK Japan Broadcasting Corporation Science and Technology Research Laboratories |
Fujii M.,NHK Japan Broadcasting Corporation Science and Technology Research Laboratories
IEICE Transactions on Information and Systems | Year: 2011
Speeded up robust features (SURF) can detect scale- and rotation-invariant features at high speed by relying on integral images for image convolutions. However, since the number of image convolutions greatly increases in proportion to the image size, another method for reducing the time for detecting features is required. In this letter, we propose a method, called ordinal convolution, of reducing the number of image convolutions for fast feature detection in SURF and compare it with a previous method based on sparse sampling. © 2011 The Institute of Electronics, Information and Communication Engineers.