Naoya Wada

Dwarf bamboos affect the regeneration of zoochorous trees by providing habitats to acorn-feeding rodents

The effects of dwarf bamboos (Sasa spp.) on the regeneration of trees in a natural hardwood forest were studied by analysing the spatial dispersion of seedlings and saplings of anemochores (Acer palmatum var. matsumurae, Fraxinus lanuginosa, and Carpinus laxiflora) and zoochores (Quercus mongolica var. grosseserrata and Q. serrata). Relative photosynthetic photon flux density at 10 cm above ground was significantly correlated with the coverage of dwarf bamboos (r=0.661, P<0.001). Seedlings were abundant and were randomly distributed in the anemochores, other than the shade-intolerant species C. laxiflora which was significantly more sparse in sites with dense Sasa than in sites where Sasa was rare. Distribution of saplings was also random in the shadetolerant anemochores A. palmatum var. matsumurae and F. lanuginosa but aggregated in sites with sparse Sasa in the shade-intolerant anemochore C. laxiflora. In contrast to the anemochores, seedlings of zoochores were very few and were distributed in sites with sparse Sasa. Saplings were also aggregated and negatively correlated with Sasa cover in the shade-intolerant species Q. serrata and the tolerant species Q. mongolica var. grosseserrata. The acorns put on the forest floor in a site with dense Sasa were quickly removed by small rodents such as Apodemus speciosus and A. argenteus. Trap census of rodents revealed that those mammals prefer the dense Sasa habitat to the sparse Sasa habitat. This suggests that the dwarf bamboos strongly affect the regeneration of zoochorous trees not only by shading the seedlings but also by providing habitats to acorn-feeding small mammals.

Scalable architecture for word HMM-based speech recognition and VLSI implementation in complete system

This paper describes a scalable architecture for real-time speech recognizers based on word hidden Markov models (HMMs) that provide high recognition accuracy for word recognition tasks. However, the size of their recognition vocabulary is small because its extremely high computational costs cause long processing times. To achieve high-speed operations, we developed a VLSI system that has a scalable architecture. The architecture effectively uses parallel computations on the word HMM structure. It can reduce processing time and/or extend the word vocabulary. To explore the practicality of our architecture, we designed and evaluated a complete system recognizer, including speech analysis and noise robustness parts, on a 0.18-/spl mu/m CMOS standard cell library and field-programmable gate array. In the CMOS standard-cell implementation, the total processing time is 56.9 /spl mu/s/word at an operating frequency of 80 MHz in a single system. The recognizer gives a real-time response using an 800-word vocabulary.

Cepstral gain normalization for noise robust speech recognition

The paper describes a robust speech recognition technique which normalizes cepstral gains in order to remove effects of additive noise. We assume that the effects can be expressed by an approximate model which consists of gain and DC components in log-spectrum. Accordingly, we propose cepstral gain normalization (CGN) which normalizes the gains by means of calculating maximum and minimum values of cepstral coefficients in speech frames. The proposed method can extract noise robust features without a priori knowledge and environmental adaptation because it is applied to both training and testing data. We have evaluated recognition performance under noisy environments using the Noisex-92 database and a 100 Japanese city names task. The CGN provides improvements of recognition accuracy at various SNRs compared with combinations of conventional methods.

Japanese maple (Acer palmatum var. Matsumurae, Aceraceae) recruitment patterns: seeds, seedlings, and saplings in relation to conspecific adult neighbors.

We analyzed the spatial patterns among seeds, seedlings, saplings, and conspecific adult trees of the cool-temperate tree species Acer palmatum var. Matsumurae in a conifer-hardwood mixed forest in northern Japan, using two models that consider the influence of each adult within the neighborhood of the offspring. The results showed that recruitment patterns of each stage could be characterized and that significant shifts occur between successive stages. Sound seeds were more widely dispersed than unsound seeds; the mean dispersal distance (MDD) was 41.5 m for sound seeds, but only 12.6 m for unsound seeds. Most seedlings were located near conspecific adult trees, with a MDD of 14.3 m. Saplings, however, were more dispersed away from conspecific adult trees, with an MDD of more than 35 m. Light and gap distributions did not strongly affect the spatial distribution of the offspring; most saplings were located under nonconspecific canopies. These results suggest that the recruitment pattern of Japanese maple offspring is strongly affected by conspecific adult neighbors, rather than by light and gap distributions, with close proximity to conspecific adult trees reducing the growth and survival of seedlings during the transition to saplings.

Impact of overgrazing on seed predation by rodents in the Thar desert, northwestern India

We compared the vegetation structure, rodent density and seed loss rate between protected and disturbed sites affected from grazing by cattle, goats and sheep, in the Thar desert of India. A perennial tussocky grassLasiurus sindicus Hent. was largely dominant in the protected site, whileL. sindicus was rare and replaced by undershrub speciesAerva pseudotomentosa Blatt. & Halb. andCrotalaria burbira Buch.-Ham. in the overgrazed site. In the grazed site, plant coverage was low, but the density of rodent burrows and the frequency of rodent captures were significantly high as compared to the protected site. Corresponding with the density of desert rodents, seed predation was significantly higher in the grazed site than in the protected site. These results suggest that overgrazing by large mammals has strong effects on plant succession by altering not only the species composition and abundance of plant community, but also the habitat suitability for seed-eating rodents.

Ecological significance of the aerial seed pool of a desert lignified annual, Blepharis sindica (Acanthaceae)

Reproductive traits of a lignified annual plant, Blepharis sindica were studied in relation to the formation of an aerial seed pool on dead plants in an arid grassland in the Thar Desert of northwestern India. The dead plants remained standing on the soil surface and retained fruits for more than one year. Aerial seed pools developed about 6 cm above the ground. There were no seed pools on or below the ground surface. Only 5.7% of seeds died on dead plants because of insect predation or fungi infection during one year. Seed release was cued by rainfall, and a fraction of seeds on the aerial seed pools was released in each rainfall event. After 13 rainfall events during the monsoon season, 25% of seeds was still retained on the plants. Seed predation on the ground surface was intensive; all cones placed on the soil surface were removed within four days, and 97% of fruits were removed within 10 days. Fifty percent of seeds germinated within 3.5 h, and there were no differences in viability and time required for germination between first year seeds and older seeds. The results indicate that the aerial seed-holding on dead plants is an available way to avoid seed predation in harsh desert environments where seed predation is intense and favorable periods for growth are temporally limited and unpredictable.

Scalable architecture for word HMM-based speech recognition

This paper presents a scalable architecture for realizing real-time speech recognizers based on a word HMM (hidden Markov model). HMM-based recognition algorithms are classified into two acoustic models, i.e., phenome-level model and word-level model. The phenome-level HMM has been widely used in current speech recognition systems which permit large-sized vocabularies. Whereas the word-level HMM has been constrained to small-sized vocabularies because of extremely high computation cost in spite of excellent recognition performance. In order to overcome the shortage, we adopt the scalable architecture focused on the word HMM structure. The proposed architecture can flexibly improve recognition performance and extend word vocabularies. In addition, the computation time is hardly increasing. In order to demonstrate practical solutions, we have designed and evaluated a total system recognizer including speech analysis and noise robustness on a 0.18 /spl mu/m CMOS standard cell library. The recognition time is 35.7 /spl mu/s/word at 128 MHz operating frequency. The recognizer can achieve over middle-sized vocabularies in real-time response.

Seed dispersal and predation by small rodents on the herbaceous understory plant Symplocarpus renifolius

-We observed predation on and dispersal of seeds of the Asian skunk cabbage Symplocarpus renifolius (Araceae), a common herb species on the mesic forest floor of northern Japan. About 11% of fertile plants succeeded in setting fruits with an average of 10.6 ? SD 13.4 seeds per plant. The germination and survival rate of seeds to the seedling stage was estimated to be about 18%. The seeds deposited in skunk cabbage vegetation were removed by the Japanese wood mouse Apodemus speciosus ainu. Many of these seeds were immediately eaten by the mice, but about 15% of them were hoarded ca. 1 cm beneath the litter layer or the moss carpet. The mean dispersal distance of the hoarded seeds was 9.6 ? SD 3.8 m. In addition to A. speciosus ainu, A. argenteus and Clethrionomys rufocanus were also abundant in this forest, but A. speciosus ainu appeared most effective in scatterhoarding seeds and most active in the skunk cabbage vegetation. These results suggest that the successful establishment of seedlings of this herb species is considerably affected by the distribution of small rodents.

Size-dependent flowering behavior and heat production of a sequential hermaphrodite, Symplocarpus renifolius (Araceae)

We examined the flowering performance in a population of the protogynous perennial herb Symplocarpus renifolius (Araceae), with special consideration of plant size. Flowering of S. renifolius occurred in very early spring, soon after snow melt. The spadices generated heat throughout the pistillate (female) and bisexual phases, but heat production decreased quickly after the beginning of the staminate (male) phase. During the flowering season, the sex ratio within the population dramatically changed from a dominance of females to a dominance of males. The duration of the female phase was negatively correlated with the onset time of flowering, and the duration of the male phase was positively correlated with plant size. Larger plants began blooming earlier, produced more heat, made the transition from female to male phase more rapidly, and lasted longer as males than smaller ones. Such size-dependent flowering performance caused unidirectional pollen flow from large to small plants. The number of seeds produced per spadix was positively correlated with the duration of the female phase, although it was not correlated with plant size. However, the estimated number of seeds sired during the male phase was positively correlated with plant size. Early flowering, rapid gender change, and higher heat production of the spadices by larger plants were factors considered to promote the higher success of the male function without decreasing the success of the female function.

Direct control on modulation spectrum for noise-robust speech recognition and spectral subtraction

On the modulation spectral domain, speech components and noise components can be separated by their modulation spectral bands. Although conventional methods extract speech components using band-pass filtering on running spectrum, those extractions are affected by limitation of filtering such as stability, filter steepness, etc. Instead of it, this paper proposes modulation spectral control which realizes the feature extraction of modulation spectral bands with weighting factors on modulation spectral domain. It is free from those limitations and enables flexible extraction of specific modulation spectral bands. The combination of it and spectral subtraction shows outstanding performance and flexibility for the variety of noise in speech recognition experiments