Research on continuous speech recognition with small vocabulary based on biomimetic pattern recognition

Based on biomimetic pattern recognition theory, we proposed a novel speaker-independent continuous speech keyword-spotting algorithm. Without endpoint detection and division, we can get the minimum distance curve between continuous speech samples and every keyword-training net through the dynamic searching to the feature-extracted continuous speech. Then we can count the number of the keywords by investigating the vale-value and the numbers of the vales in the curve. Experiments of small vocabulary continuous speech with various speaking rate have got good recognition results and proved the validity of the algorithm.

Geometry property of multi-degree of freedom neurons

In this paper, we study a problem of geometric inequalities for a Multi-degree of Freedom Neurons. Some new geometric inequalities for a Multi-degree of Freedom Neurons are established. As special cases, some known inequalities are deduced.

Cognitive models in biomimetic pattern cognition

We describe a new model which is based on the concept of cognizing theory. The method identifies subsets of the data which are embedded in arbitrary oriented lower dimensional space. We definite k-mean covering, and study its property. Covering subsets of points are repeatedly sampled to construct trial geometry space of various dimensions. The sampling corresponding to the feature space having the best cognition ability between a mode near zero and the rest is selected and the data points are partitioned on the basis of the best cognition ability. The repeated sampling then continues recursively on each block of the data. We propose this algorithm based on cognition models. The experimental results for face recognition demonstrate that the correct rejection rate of the test samples excluded in the classes of training samples is very high and effective.

Cancer subtypes recognition and its gene expression profiles analysis based on geometrical learning

The accurate recognition of cancer subtypes is very significant in clinic. Especially, the DNA microarray gene expression technology is applied to diagnosing and recognizing cancer types. This paper proposed a method of that recognized cancer subtypes based on geometrical learning. Firstly, the cancer genes expression profiles data was pretreated and selected feature genes by conventional method; then the expression data of feature genes in the training samples was construed each convex hull in the high-dimensional space using training algorithm of geometrical learning, while the independent test set was tested by the recognition algorithm of geometrical learning. The method was applied to the human acute leukemia gene expression data. The accuracy rate reached to 100%. The experiments have proved its efficiency and feasibility.

Phylogeny of SARS-CoV based on high-dimensional information geometry

We proposed a novel methodology, which firstly, extracting features from species' complete genome data, using k-tuple, followed by studying the evolutionary relationship between SARS-CoV and other coronavirus species using the method, called "High-dimensional information geometry". We also used the mothod, namely "caculating of Minimum Spanning Tree", to construct the Phyligenetic tree of the coronavirus. From construction of the unrooted phylogenetic tree, we found out that the evolution distance between SARS-CoV and other coronavirus species is comparatively far. The tree accurately rebuilt the three groups of other coronavirus. We also validated the assertion from other literatures that SARS-CoV is similar to the coronavirus species in Group I.

Determining of the delay time for a heating ventilating and air-conditioning plant using multi-weights neurons approach

This paper presents an multi weights neurons approach to determine the delay time for a Heating ventilating and air-conditioning (HVAC) plan to respond to control actions. The multi weights neurons is a fully connected four-layer network. An acceleration technique was used to improve the general delta rule for the learning process. Experimental data for heating and cooling modes were used with both the multi weights neurons and a traditional mathematical method to determine the delay time. The results show that multi weights neurons can be used effectively determining the delay time for HVAC systems.

Optical characteristics of DBR with inhomogeneous graded interfaces

Optical properties of Al0.9Ga0.1As/Al gamma Ga1-gamma As/GaAs/Al chi Ga1-chi As DBR with inhomogeneous graded interfaces has been investigated by using characteristic matrix method. The refractive index model and the analytic characteristic matrix of graded interfaces are obtained. The reflectance spectrum and the reflective phase shift are calculated for GaAs/Al-0.9 Ga-0.1 As DBR and graded interfaces DBR by using characteristic matrix method. The effect of graded interfaces on the optical properties of DBR is discussed. The result shows an extra graded phase matching layer must he added in front of the graded interfaces DBR to fulfil the conditions of phase matching at central wavelength. The accurate thickness of phase matching layer is calculated by optical thickness approximation method.

Influence of GaAsP insertion layers on performance of InGaAsP/InGaP/AlGaAs quantum-well laser

We report on the use of very thin GaAsP insertion layers to improve the performance of an InGaAsP/InGaP/AlGaAs single quantum-well laser structure grown by metal organic chemical vapour deposition. Compared to the non-insertion structure, the full width at half maximum of photoluminescence spectrum of the insertion structure measured at room temperature is decreased from 47 to 38 nm indicating sharper interfaces. X-ray diffraction shows that the GaAsP insertion layers between AlGaAs and InGaP compensates for the compressive strain to improve the total interface. The laser performance of the insertion structure is significantly improved as compared with the counterpart without the insertion layers. The threshold current is decreased from 560 to 450mA while the slope efficiency is increased from 0.61 to 0.7W/A and the output power is increased from 370 to 940mW. The slope efficiency improved is very high for the devices without coated facets. The improved laser performance is attributed to the suppression of indium carry-over due to the use of the GaAsP insertion layers.

Influences of quantum noises on direct-modulated properties of 1.3-mu m InGaAsP/InP laser diodes

Due to the zero dispersion point at 1.3-mu m in optical fibres, 1.3-mu m InGaAsP/InP laser diodes have become main light sources in fibre communication systems recently. In fluences of quantum noises on direct-modulated properties of single-mode 1.3-mu m InGaAsP/InP laser diodes are investigated in this article. Considering the carrier and photon noises and the cross-correlation between the two noises, the power spectrum of the photon density and the signal-to-noise ratio (SNR) of the direct-modulated single-mode laser system are calculated using the linear approximation method. We find that the stochastic resonance (SR) always appears in the dependence of the SNR on the bias current density, and is strongly affected by the cross-correlation coeffcient between the carrier and photon noises, the frequency of modulation signal, and the photon lifetime in the laser cavity. Hence, it is promising to use the SR mechanism to enhance the SNR of direct-modulated InGaAsP/InP laser diodes and improve the quality of optical fibre communication systems.

Study on text-dependent speaker recognition based on Biomimetic Pattern Recognition

We studied the application of Biomimetic Pattern Recognition to speaker recognition. A speaker recognition neural network using network matching degree as criterion is proposed. It has been used in the system of text-dependent speaker recognition. Experimental results show that good effect could be obtained even with lesser samples. Furthermore, the misrecognition caused by untrained speakers occurring in testing could be controlled effectively. In addition, the basic idea "cognition" of Biomimetic Pattern Recognition results in no requirement of retraining the old system for enrolling new speakers.

High-Dimensional Space Geometrical Informatics and its applications to image restoration

With a view to solve the problems in modern information science, we put forward a new subject named High-Dimensional Space Geometrical Informatics (HDSGI). It builds a bridge between information science and point distribution analysis in high-dimensional space. A good many experimental results certified the correctness and availability of the theory of HDSGI. The proposed method for image restoration is an instance of its application in signal processing. Using an iterative "further blurring-debluring-further blurring" algorithm, the deblured image could be obtained.

Anti-Stokes photoluminescence in ZnO microcrystal

Low temperature (10 K) strong anti-Stokes photoluminescence (ASPL) of ZnO microcrystal excited by low power cw 532 nm laser is reported here. Energy upconversion of 1.1 eV is obtained in our experiment with no conventional nonlinear effect. Through the study of the normal photoluminescence and temperature dependence of ASPL we conclude that the green band luminescence in ZnO is related to deep donor to valance band transition. Using the two-step two-photon absorption model, we provide a plausible mechanism leading to the ASPL phenomenon in our experiment. (c) 2006 American Institute of Physics.

Continuous speech recognition based on ICA and geometrical learning

We investigate the use of independent component analysis (ICA) for speech feature extraction in digits speech recognition systems. We observe that this may be true for recognition tasks based on Geometrical Learning with little training data. In contrast to image processing, phase information is not essential for digits speech recognition. We therefore propose a new scheme that shows how the phase sensitivity can be removed by using an analytical description of the ICA-adapted basis functions. Furthermore, since the basis functions are not shift invariant, we extend the method to include a frequency-based ICA stage that removes redundant time shift information. The digits speech recognition results show promising accuracy. Experiments show that the method based on ICA and Geometrical Learning outperforms HMM in a different number of training samples.