Takio KURITA Toshiharu TAGUCHI
This paper presents a modification of kernel-based Fisher discriminant analysis (FDA) to design one-class classifier for face detection. In face detection, it is reasonable to assume "face" images to cluster in certain way, but "non face" images usually do not cluster since different kinds of images are included. It is difficult to model "non face" images as a single distribution in the discriminant space constructed by the usual two-class FDA. Also the dimension of the discriminant space constructed by the usual two-class FDA is bounded by 1. This means that we can not obtain higher dimensional discriminant space. To overcome these drawbacks of the usual two-class FDA, the discriminant criterion of FDA is modified such that the trace of covariance matrix of "face" class is minimized and the sum of squared errors between the average vector of "face" class and feature vectors of "non face" images are maximized. By this modification a higher dimensional discriminant space can be obtained. Experiments are conducted on "face" and "non face" classification using face images gathered from the available face databases and many face images on the Web. The results show that the proposed method can outperform the support vector machine (SVM). A close relationship between the proposed kernel-based FDA and kernel-based Principal Component Analysis (PCA) is also discussed.
Amaro LIMA Heiga ZEN Yoshihiko NANKAKU Keiichi TOKUDA Tadashi KITAMURA Fernando G. RESENDE
This paper presents an analysis of the applicability of Sparse Kernel Principal Component Analysis (SKPCA) for feature extraction in speech recognition, as well as, a proposed approach to make the SKPCA technique realizable for a large amount of training data, which is an usual context in speech recognition systems. Although the KPCA (Kernel Principal Component Analysis) has proved to be an efficient technique for being applied to speech recognition, it has the disadvantage of requiring training data reduction, when its amount is excessively large. This data reduction is important to avoid computational unfeasibility and/or an extremely high computational burden related to the feature representation step of the training and the test data evaluations. The standard approach to perform this data reduction is to randomly choose frames from the original data set, which does not necessarily provide a good statistical representation of the original data set. In order to solve this problem a likelihood related re-estimation procedure was applied to the KPCA framework, thus creating the SKPCA, which nevertheless is not realizable for large training databases. The proposed approach consists in clustering the training data and applying to these clusters a SKPCA like data reduction technique generating the reduced data clusters. These reduced data clusters are merged and reduced in a recursive procedure until just one cluster is obtained, making the SKPCA approach realizable for a large amount of training data. The experimental results show the efficiency of SKPCA technique with the proposed approach over the KPCA with the standard sparse solution using randomly chosen frames and the standard feature extraction techniques.
Satoshi UKAI Tomoya TAKATANI Hiroshi SARUWATARI Kiyohiro SHIKANO Ryo MUKAI Hiroshi SAWADA
In this paper, single-input multiple-output (SIMO)-model-based blind source separation (BSS) is addressed, where unknown mixed source signals are detected at microphones, and can be separated, not into monaural source signals but into SIMO-model-based signals from independent sources as they are at the microphones. This technique is highly applicable to high-fidelity signal processing such as binaural signal processing. First, we provide an experimental comparison between two kinds of SIMO-model-based BSS methods, namely, conventional frequency-domain ICA with projection-back processing (FDICA-PB), and SIMO-ICA which was recently proposed by the authors. Secondly, we propose a new combination technique of the FDICA-PB and SIMO-ICA, which can achieve a higher separation performance than the two methods. The experimental results reveal that the accuracy of the separated SIMO signals in the simple SIMO-ICA is inferior to that of the signals obtained by FDICA-PB under low-quality initial value conditions, but the proposed combination technique can outperform both simple FDICA-PB and SIMO-ICA.
Takahiro MURAKAMI Tetsuya HOYA Yoshihisa ISHIDA
This paper presents a novel algorithm for spectral subtraction (SS). The method is derived from a relation between the spectrum obtained by the discrete Fourier transform (DFT) and that by a subspace decomposition method. By using the relation, it is shown that a noise reduction algorithm based on subspace decomposition is led to an SS method in which noise components in an observed signal are eliminated by subtracting variance of noise process in the frequency domain. Moreover, it is shown that the method can significantly reduce computational complexity in comparison with the method based on the standard subspace decomposition. In a similar manner to the conventional SS methods, our method also exploits the variance of noise process estimated from a preceding segment where speech is absent, whereas the noise is present. In order to more reliably detect such non-speech segments, a novel robust voice activity detector (VAD) is then proposed. The VAD utilizes the spread of eigenvalues of an autocorrelation matrix corresponding to the observed signal. Simulation results show that the proposed method yields an improved enhancement quality in comparison with the conventional SS based schemes.
Toshifumi MORIYAMA Seiho URATSUKA Toshihiko UMEHARA Hideo MAENO Makoto SATAKE Akitsugu NADAI Kazuki NAKAMURA
This paper describes a polarimetric feature extraction method from urban areas using the POLSAR (Polarimetric Synthetic Aperture Radar) data. The scattering characteristic of urban areas is different from that of natural distributed areas. The main point of difference is polarimetric correlation coefficient, because urban areas do not satisfy property of azimuth symmetry, Shh = Shv = 0. The decomposition technique based on azimuth symmetry can not be applied to urban areas. We propose a new model fit suitable for urban areas. The proposed model fit consists of odd-bounce, even-bounce and cross scattering models. These scattering models can represent the polarimetric backscatter from urban areas, and satisfy Shh 0 and Shv 0. In addition, the combination with the proposed model fit and the three component scattering model suited for natural distributed areas is examined. It is possible to apply the combined technique to POLSAR data which includes both urban areas and natural distributed areas. The combined technique is used for feature extraction of actual X-band POLSAR data acquired by Pi-SAR. It is shown that the proposed model fit is useful to extract polarimetric features from urban areas.
We propose a novel shadow texture generation method with linear processing time using a shadow depth buffer (SZ-Buffer). We also present a method that achieves further speedup using temporal coherence. If the transition between dynamic and static state is not frequent, depth values of static objects does not vary significantly. So we can reuse the depth value for static objects and render only dynamic objects.
Xuan Nam TRAN Tetsuki TANIGUCHI Yoshio KARASAWA
In this paper, we propose a spatio-temporal equalizer for the space-time block coded transmission over the frequency selective fading channels with the presence of co-channel interference (CCI). The proposed equalizer, based on the tapped delay line adaptive array (TDLAA), performs signal equalization and CCI suppression simultaneously using the minimum mean square error (MMSE) method. It is to show that our scheme outperforms the previous two-stage combined adaptive antenna and delayed decision feedback sequence estimator (DDFSE) approach. We also show that performance can be further improved if the synchronization between the preceding and delayed paths is achieved.
In a ubiquitous computing environment, people are surrounded by hundreds of mobile or embedded computers each of which may be used to support one or more user applications due to limitations in their individual computational capabilities. We need an approach to coordinating heterogeneous computers that acts as a virtual computer around a mobile and ubiquitous computing environment and supports various applications beyond the capabilities of single computers. This paper presents a framework for building and aggregating distributed applications from one or more mobile components that can be dynamically deployed at mobile or stationary computers during the execution of the application. Since the approach involves mobile-transparent communications between components and component relocation semantics, it enables a federation of components to adapt its structure and deployment on multiple computers whose computational resources, such as input and output devices, can satisfy the requirement of the components in a self-organized manner. This paper also describes a prototype implementation of the approach and its application.
Nobukazu TAKAI Shigetaka TAKAGI Nobuo FUJII
This paper proposes a rail-to-rail OTA. By adding a signal decomposing circuit at the input of given OTAs that have a limited input voltage range, a rail-to-rail OTA is obtained. Each decomposed input voltage signal is converted to a current signal by an OTA and each output current of OTAs is summed to obtain a linear output signal. Since the input signal is decomposed into small magnitude voltage signals, the OTAs used to the voltage-current conversion do not require a wide input-range and any OTA can be used to realize a rail-to-rail input voltage range OTA. HSPICE simulations are performed to verify the validity of the proposed method.
We propose a novel approach based on wavelet decomposition for progressive full spectral rendering. In the fourth progressive stage, our method renders an image that is 95% similar to the final non-progressive approach but requires less than 70% of the execution time. The quality of the rendered image is visually plausible that is indistinguishable from that of the non-progressive method. Our approach is graceful, efficient, progressive, and flexible for full spectral rendering.
Kazuhiko USHIO Hideaki FUJIMOTO
We show that the necessary and sufficient condition for the existence of a balanced quatrefoil decomposition of the complete multigraph λKn is n 9 and λ(n - 1) 0 (mod 24). Decomposition algorithms are also given.
Mitsuyoshi KISHIHARA Kuniyoshi YAMANE Isao OHTA Tadashi KAWAI
This paper treats multi-way microstrip power dividers composed of multi-step, multi-furcation, and corners. Since the design procedure is founded on the planar circuit approach in combination with the segmentation method, optimization of the circuit configuration can be performed in a reasonable short computation time when applying the Powell's optimization algorithm. Actually, broadband 3- and 4-way power dividers with mitered bends are designed, and fractional bandwidths of about 90% and 100% are realized for the power-split imbalance less than 0.2 dB and the return loss better than -20 dB, respectively. The validity of the design results is confirmed by an EM-simulator (HFSS) and experiments.
This paper presents a design procedure of a directional coupler consisting of a twofold symmetric four-port circuit with four identical matching networks at each port. The intrinsic power-split ratio and the equivalent admittance of the directional coupler are formularized in terms of the eigenadmittances of the original four-port without the matching networks. These formulas are useful for judgment on the realizability of a directional coupler in a given circuit structure and for design of the matching networks. Actually, the present procedure is applied to designing various quadrature hybrids and directional couplers, and its practical usefulness as well as several new circuit structures are demonstrated.
Hiroyuki KAWAI Kenichi HIGUCHI Noriyuki MAEDA Mamoru SAWAHASHI Takumi ITO Yoshikazu KAKURA Akihisa USHIROKAWA Hiroyuki SEKI
This paper proposes likelihood function generation of complexity-reduced Maximum Likelihood Detection with QR Decomposition and M-algorithm (QRM-MLD) suitable for soft-decision Turbo decoding and investigates the throughput performance using QRM-MLD with the proposed likelihood function in multipath Rayleigh fading channels for Orthogonal Frequency and Code Division Multiplexing (OFCDM) multiple-input multiple-output (MIMO) multiplexing. Simulation results show that by using the proposed likelihood function generation scheme for soft-decision Turbo decoding following QRM-MLD in 4-by-4 MIMO multiplexing, the required average received signal energy per bit-to-noise power spectrum density ratio (Eb/N0) at the average block error rate (BLER) of 10-2 at a 1-Gbps data rate is significantly reduced compared to that using hard-decision decoding in OFCDM access with 16 QAM modulation, the coding rate of 8/9, and 8-code multiplexing with a spreading factor of 8 assuming a 100-MHz bandwidth. Furthermore, we show that by employing QRM-MLD associated with soft-decision Turbo decoding for 4-by-4 MIMO multiplexing, the throughput values of 500 Mbps and 1 Gbps are achieved at the average received Eb/N0 of approximately 4.5 and 9.3 dB by QPSK with the coding rate of R = 8/9 and 16QAM with R = 8/9, respectively, for OFCDM access assuming a 100-MHz bandwidth in a twelve-path Rayleigh fading channel.
Nari TANABE Toshihiro FURUKAWA Kohichi SAKANIWA Shigeo TSUJII
We propose a practical blind channel identification algorithm based on the principal component analysis. The algorithm estimates (1) the channel order, (2) the noise variance, and then identifies (3) the channel impulse response, from the autocorrelation of the channel output signal without using the eigenvalue and singular-value decomposition. The special features of the proposed algorithm are (1) practical method to find the channel order and (2) reduction of computational complexity. Numerical examples show the effectiveness of the proposed algorithm.
Juntae YOON Seonho KIM Hae-Chang RIM
This paper presents a method for improving the performance of syntactic analysis by using accurate temporal expression processing. Temporal expression causes parsing errors due to its syntactic duality, but its resolution is not trivial since the syntactic role of temporal expression is understandable in the context. In our work, syntactic functions of temporal words are decisively identified based on local contexts of individual temporal words acquired from a large corpus, which are represented by a finite state method. Experimental results show how the proposed method, incorporated with parsing, improves the accuracy and efficiency of the syntactic analysis.
Amaro LIMA Heiga ZEN Yoshihiko NANKAKU Chiyomi MIYAJIMA Keiichi TOKUDA Tadashi KITAMURA
This paper describes an approach to feature extraction in speech recognition systems using kernel principal component analysis (KPCA). This approach represents speech features as the projection of the mel-cepstral coefficients mapped into a feature space via a non-linear mapping onto the principal components. The non-linear mapping is implicitly performed using the kernel-trick, which is a useful way of not mapping the input space into a feature space explicitly, making this mapping computationally feasible. It is shown that the application of dynamic (Δ) and acceleration (ΔΔ) coefficients, before and/or after the KPCA feature extraction procedure, is essential in order to obtain higher classification performance. Better results were obtained by using this approach when compared to the standard technique.
Yoshitaka UKAWA Toshimitsu USHIO Masakazu ADACHI Shigemasa TAKAI
In this paper, we propose a formal method for detection of three automation surprises in human-machine interaction; a mode confusion, a refusal state, and a blocking state. The mode confusion arises when a machine is in a different mode from that anticipated by the user, and is the most famous automation surprise. The refusal state is a situation that the machine does not respond to a command the user executes. The blocking state is a situation where an internal event occurs, leading to change of an interface the user does not know. In order to detect these phenomena, we propose a composite model in which a machine and a user model evolve concurrently. We show that the detection of these phenomena in human-machine interaction can be reduced to a reachability problem in the composite model.
Kazuhiko USHIO Hideaki FUJIMOTO
We show that the necessary and sufficient condition for the existence of a balanced bowtie decomposition of the symmetric complete multi-digraph is n 5 and λ(n-1) 0 (mod 6). Decomposition algorithms are also given.
Parallel concatenated convolutional codes, turbo codes, are very attractive scheme at a point of view of an error probability performance. An bit error rate (BER) evaluation for turbo codes is done by a uniform interleaver bound calculation and/or a computer simulation. The former is calculated under the assumption of uniform interleaver, and is only effective for an BER evaluation with a pseudo random interleaver. The latter dose not have any interleaver restrictions. However, for a very low BER evaluation, it takes enormous simulation time. In this paper, a new error probability evaluation method for turbo codes is proposed. It is based on the error event simulation method. For each evaluation for the predetermined error sequence, importance sampling, which is one of the fast simulation methods, is applied. To prove the effectiveness of the proposed method, numerical examples are shown. The proposed method well approximates the BER at the error floor region. Under the same accuracy, the IS estimation time at BER = 10-7 is reduced to 1/6358 of the ordinary Monte-Carlo simulation time.