Continuous speech recognition for the TIMIT database using neural networks

Four types of neural networks which have previously been established for speech recognition and tested on a small, seven-speaker, 100-sentence database are applied to the TIMIT database. The networks are a recurrent network phoneme recognizer, a modified Kanerva model morph recognizer, a compositional representation phoneme-to-word recognizer, and a modified Kanerva model morph-to-word recognizer. The major result is for the recurrent net, giving a phoneme recognition accuracy of 57% from the si and sx sentences. The Kanerva morph recognizer achieves 66.2% accuracy for a small subset of the sa and sx sentences. The results for the word recognizers are incomplete.

MMI training for continuous phoneme recognition on the TIMIT database

This paper reports our experiences with a phoneme recognition system for the TIMIT database which uses multiple mixture continuous density monophone HMMs trained using MMI. A comprehensive set of results are presented comparing the ML and MMI training criteria for both diagonal and full covariance models. These results using simple monophone HMMs show clear performance gains achieved by MMI training, and are comparable to the best reported by others including those which use context-dependent models. In addition, the paper discusses a number of performance and implementation issues which are crucial to successful MMI training.

Physical variables, nutrient status and primary productivity of Kyoga Basin lakes of Lemwa, Omunuo, Gawa, Adois and Owapet

The review report on Kyoga basin lakes (NAFIRRI 2007) described Kyoga basin lakes as important natural resource for the communities within the basin and the surrounding areas. Fisheries of the basin provide a source of protein, income, and employment to generally poor communities in the area. The lakes also generate revenue to the local Governments within the catchment. This indicates that the fisheries of Kyoga basin lakes are a key instrument in poverty eradication and food security. The lakes also act as a source of water for domestic, agricultural and transport purposes. Some of the Kyoga small lakes harbour fish species, which have disappeared from the main lakes Victoria and Kyoga and are therefore important for biodiversity conservation

Joint modelling of voicing label and continuous F0 for HMM based speech synthesis

Fundamental frequency, or F0 is critical for high quality speech synthesis in HMM based speech synthesis. Traditionally, F0 values are considered to depend on a binary voicing decision such that they are continuous in voiced regions and undefined in unvoiced regions. Multi-space distribution HMM (MSDHMM) has been used for modelling the discontinuous F0. Recently, a continuous F0 modelling framework has been proposed and shown to be effective, where continuous F0 observations are assumed to always exist and voicing labels are explicitly modelled by an independent stream. In this paper, a refined continuous F0 modelling approach is proposed. Here, F0 values are assumed to be dependent on voicing labels and both are jointly modelled in a single stream. Due to the enforced dependency, the new method can effectively reduce the voicing classification error. Subjective listening tests also demonstrate that the new approach can yield significant improvements on the naturalness of the synthesised speech. A dynamic random unvoiced F0 generation method is also investigated. Experiments show that it has significant effect on the quality of synthesised speech. © 2011 IEEE.

Structured discriminative models for noise robust continuous speech recognition

Recently there has been interest in structured discriminative models for speech recognition. In these models sentence posteriors are directly modelled, given a set of features extracted from the observation sequence, and hypothesised word sequence. In previous work these discriminative models have been combined with features derived from generative models for noise-robust speech recognition for continuous digits. This paper extends this work to medium to large vocabulary tasks. The form of the score-space extracted using the generative models, and parameter tying of the discriminative model, are both discussed. Update formulae for both conditional maximum likelihood and minimum Bayes' risk training are described. Experimental results are presented on small and medium to large vocabulary noise-corrupted speech recognition tasks: AURORA 2 and 4. © 2011 IEEE.

Enhanced poisson sum representation for alpha-stable processes

In this paper we present Poisson sum series representations for α-stable (αS) random variables and a-stable processes, in particular concentrating on continuous-time autoregressive (CAR) models driven by α-stable Lévy processes. Our representations aim to provide a conditionally Gaussian framework, which will allow parameter estimation using Rao-Blackwellised versions of state of the art Bayesian computational methods such as particle filters and Markov chain Monte Carlo (MCMC). To overcome the issues due to truncation of the series, novel residual approximations are developed. Simulations demonstrate the potential of these Poisson sum representations for inference in otherwise intractable α-stable models. © 2011 IEEE.

Investigations on wooden floating materials: the loss of buoyancy under continuous immersion in water and the rate of its redemption by drying

A wooden fishing float under immersion in water for long periods is liable to absorb water, the quantity of water absorbed possibly being dependent upon the physical factors like the specific gravity and the inherent property of the material, the time of soaking and the pressure acting on it. Consequently a wooden float is likely to become heavy and loss its original buoyancy. However, when the float is removed from water and dried, the lost buoyancy is regained on complete drying. The present paper is an attempt to elucidate these two important characteristics of some of the chief wooden floating materials used on the West Coast of India.

Discriminant analysis of selected variables in the adoption of fish curing

Discriminant functions were worked out for adoption or non-adoption of five improved practices in fish curing. Four variables measured quantitatively formed the basis for discrimination. In four out of five equations, the selected variables were found to discriminate significantly between the adopters and non-adopters.

Geographic variation of Chevrier's field mouse (Apodemus chevrieri) (Milne-Edwards, 1868) (Muridae : Murinae) from southwestern China based on cranial morphometric variables

Geographic variation of Chevrier's field mouse (Apodemus chevrieri) (Milne-Edwards, 1868) (Muridae: Murinae) from southwestern China based on cranial morphometric variables. Zoological Studies 47(4): 393-401. A sample of 134 specimens of Apodemus chevrieri was investigated in the present study. Individuals were divided into male and female groups, and these were respectively subjected to multivariate analysis. Results indicated that 3 geographic populations of A. chevrieri inhabit southwestern China: a Sichuan population in western Sichuan Province; a northwestern Yunnan population ranging from northwestern Yunnan Province eastward to southern Sichuan Province; and a central Yunnan population in central Yunnan Province. In addition, a coefficient of difference analysis was performed among these 3 geographic populations. The results suggested that these 3 geographical populations of A. chevrieri belonged to 2 subspecies. Furthermore, we discuss the relationships of the subspecific differentiation of A. chevneri with changes in latitude in southwestern China.

Geographic variation of the Perny's Long-nosed squirrels (Dremomys pernyi) (Milne-Edwards, 1867) (Rodentia : Sciuridae) from southwestern China based on cranial morphometric variables

A sample of 114 specimens of Dremomys pernyi was investigated, 73 of which had intact skulls and were subjected to multivariate, coefficient of difference (C. D.), and cluster analyses. Results indicate that 4 subspecies (groups) of Dremomys pernyi inhabi

Geographic variation of the Anderson's Niviventer (Niviventer andersoni) (Thomas, 1911) (Rodentia: Muridae) of two new subspecies in China verified with cranial morphometric variables and pelage characteristics

A total of 66 specimens of Niviventer andersoni with intact skulls was investigated on pelage characteristics and cranial morphometric variables. The data were subjected to principal component analyses as well as to discriminant analyses, and measurement