- 著者
- Sato Kaede Matsuno Kohei Arima Daichi Abe Yoshiyuki Yamaguchi Atsushi
- 出版者
- SpringerOpen
- 雑誌
- Zoological Studies (ISSN:1810522X)
- 巻号頁・発行日
- vol.54, no.1, 2015-01-16
- 被引用文献数
- 18 1
Background An optical plankton counter (OPC) was used to examine spatial and temporal changes in the zooplankton size spectra in the neighboring waters of Japan from May to August 2011. Results Based on the zooplankton biovolume of equivalent spherical diameter (ESD) in 45 bins for every 0.1 mm between 0.5 and 5.0 mm, a Bray-Curtis cluster analysis classified the zooplankton communities into six groups. The geographical distribution of each group varied from each of the others. Groups with a dominance of 4 to 5 mm ESD were observed in northern marginal seas (northern Japan Sea and Okhotsk Sea), while the least biovolume with a dominance of a small-size class (0.5 to 1 mm) was observed for the Kuroshio extension. Temporal changes were observed along the 155° E line, i.e., a high biovolume group dominated by 2 to 3 mm ESD during May shifted to other size spectra groups during July to August. These temporal changes were caused by the seasonal vertical descent of dominant large Neocalanus copepods during July to August. As a specific characteristic of the normalized biomass size spectra (NBSS), the slope of NBSS was moderate (−0.90) for the Neocalanus dominant spring group but was at −1.11 to −1.24 for the other groups. Theoretically, the slope of the NBSS of the stable marine ecosystem is known to settle at approximately −1. Conclusions Based on the analysis by OPC, zooplankton size spectra in the neighboring waters of Japan were separated into six groups. Most groups had −1.11 to −1.24 NBSS slopes, which were slightly higher than the theoretical value (−1). However, one group had a moderate slope of NBSS (−0.90) caused by the dominance of large Neocalanus copepods.
- 著者
- Mimura Masato Sakai Shinsuke Kawahara Tatsuya
- 出版者
- SpringerOpen
- 雑誌
- EURASIP Journal on Advances in Signal Processing (ISSN:16876172)
- 巻号頁・発行日
- vol.2015, no.1, 2015-07-23
- 被引用文献数
- 10
We propose an approach to reverberant speech recognition adopting deep learning in the front-end as well as b a c k-e n d o f a r e v e r b e r a n t s p e e c h r e c o g n i t i o n s y s t e m, a n d a n o v e l m e t h o d t o i m p r o v e t h e d e r e v e r b e r a t i o n p e r f o r m a n c e of the front-end network using phone-class information. At the front-end, we adopt a deep autoencoder (DAE) for enhancing the speech feature parameters, and speech recognition is performed in the back-end using DNN-HMM acoustic models trained on multi-condition data. The system was evaluated through the ASR task in the Reverb Challenge 2014. The DNN-HMM system trained on the multi-condition training set achieved a conspicuously higher word accuracy compared to the MLLR-adapted GMM-HMM system trained on the same data. Furthermore, feature enhancement with the deep autoencoder contributed to the improvement of recognition accuracy especially in the more adverse conditions. While the mapping between reverberant and clean speech in DAE-based dereverberation is conventionally conducted only with the acoustic information, we presume the mapping is also dependent on the phone information. Therefore, we propose a new scheme (pDAE), which augments a phone-class feature to the standard acoustic features as input. Two types of the phone-class feature are investigated. One is the hard recognition result of monophones, and the other is a soft representation derived from the posterior outputs of monophone DNN. The augmented feature in either type results in a significant improvement (7–8 % relative) from the standard DAE.