Xiaomin Mu

Recognizing human emotional state based on the phase information of the two dimensional fractional Fourier transform

Over the last decade, automatic facial expression analysis has become an active research area which finds potential applications in fields such as more engaging human-computer interaction, multimedia information analysis and retrieval, biometrics for security and surveillance, entertainment and e-health. In this paper, we explore a new class of visual features for recognizing human emotion states from. It performs feature extraction by using the method of two dimensional fractional Fourier transform (2D-FrFT). As a generalization of Fourier transform, the 2D-FrFT contains the time-frequency information of the signal at the same time, and is a new and powerful tool for time-frequency analysis. In particular, features are extracted from the phase parts of the 2DFrFT, and used to train the Fishers Linear Discriminant Analysis (FLDA) classifiers for human emotion recognition. Preliminary experiments show that the proposed 2D-FrFT features yield promising results in visual human emotion recognition. More importantly, the 2D-FrFT and the one dimensional fractional Fourier transform provide a natural, versatile and powerful platform for general audiovisual signal processing tasks.

DOA Estimation of Coherent Wideband LFM Signals Based on Fractional Fourier Transform

Referring to the coherent wideband linear frequency modulation (LFM) signals, direction of arrival (DOA) estimation based on fractional Fourier transform (FRFT) is proposed. In this method, the observed signals are first transformed to fractional Fourier (FRF) domain, using the superior time-frequency localization, time-variant direction matrix is changed into time-invariant one. By dint of the forward/backward (FB) spatial smoothing techniques, the corresponding correlation matrix is constructed. Then the DOA of multicomponent coherent wideband LFM signals can be estimated by the traditional MUSIC algorithm simultaneously. Simulation verifies the effectiveness of the novel method

Recognizing Human Emotional State Based on the 2D-FrFT and FLDA

Computer recognition of human emotional states is an important component for efficient human-computer interaction. In this paper we explore an approach for recognition of human emotion from the visual information. We perform feature selection by using The Two Dimensions Fractional Fourier Transform. As a generalization of Fourier transform, the Two Dimensions Fractional Fourier Transform (2D-FrFT) contains simultaneity the time-frequency information of the signal, and is considered as a new tool for time-frequency analysis, especially in the area of image recognition. In this paper we use the amplitude of the 2D-FrFT of an image as the feature of the images to train the FLDA multiclassifier for human emotion recognition. Simulation result shows that this method of the features extraction and classifier has high recognition rate.

A new method for peak-to-average power ratio reduction in MIMO-OFDM system

An improved method based on cross-antenna rotation and inversion (CARI) is presented for MIMO-OFDM peak-to-average power ratio (PAPR) reduction. Firstly, an optimal scheme is proposed for PAPR reduction, by employing the degrees of freedom of multiple antennas. Then two sub-optimal schemes are proposed to reduce computational complexity by transforming the subblock successively and setting a threshold. Simulation results have shown that the proposed suboptimal scheme, termed adaptive subblock successive transform (ASST) slightly outperforms the successive suboptimal CARI (SS-CARI) scheme but with remarkably lower computational complexity.

Decision aided compensation of residual frequency offset for MIMO-OFDM systems with nonlinear channel

In this paper, we propose a new approach to compensate the residual frequency offset (RFO) in multi-input multi-output orthogonal frequency division multiplexing (MIMO-OFDM) system with nonlinear channel working in the burst mode. The proposed approach consists of two stages. Firstly a decision aided method is proposed to eliminate the nonlinearity introduced by high power transmit amplifier (HPA). Then a new decision aided approach is employed to achieve the RFO compensation on the nonlinearity-free symbols. The effectiveness of the proposed approach has been verified by computer simulations.

Recognizing facial expressions based on Gabor filter selection

Recognition of human emotional state is an important component for efficient human-computer interaction. In this paper a method of Gabor filter selection for facial expression recognition is investigated. We first preprocess facial images based on affine transform to normalize the faces. Then the using of a separability judgment is proposed to evaluate the separability of different Gabor filters, and only use those filters that can better separate different expressions. In the recognition process a PCA and FLDA multiclassifier scheme is used. The experiment result shows that the introducing of Gabor filter selection can not only reduce the dimension of feature space but also reduce the computation complexity significantly, while retaining high recognition rate of above 93%.

2-D DOA Estimation of Wideband LFM Signal in Fractional Fourier Domain

In allusion to non-stationary wideband LFM signals, an algorithm of direction-of-arrival (DOA) estimation based on fractional Fourier transform (FRFT) is proposed. In this method, the observed signals are first transformed to fractional Fourier (FRF) domain, and by the unique concentration signature of signal, the FRF domain steering vector is defined, then, in the corresponding FRF domain, the two-dimension (2-D) angles estimating of the LFM signal is implemented by using DOA matrix algorithm. Theoretical analyses and simulation results show that, this algorithm can obtain good angle estimation performance with a low computation complexity

Decision aided joint compensation of clipping noise and nonlinearity for MIMO-OFDM systems

In this paper, we propose a new iterative approach to compensate the nonlinearity and clipping noise in multi-input multi-output orthogonal frequency division multiplexing (MlMO-OFDM) system with nonlinear channel. The proposed approach consists of two stages. Firstly a decision aided method is proposed to eliminate the nonlinearity introduced by high power transmit amplifier (HPA). Then this improved decision observations are employed to achieve the clipping noise compensation. The effectiveness of the proposed approach has been verified by computer simulations.

Parameter Estimation of LFM Signal in the Fractional Fourier Domain via Curve-Fitting Optimization Technique

In many ways about solving the large workload estimation of LFM signal in the FrF Domain, the 2D peak searching algorithm is much more complex. This paper presents an FrFT modulus detector via MLS curve-fitting optimization technique, which can simplify the 2D peak searching to a problem of 1D curve fitting. And for multi-component signals, we further introduce a Gaussian mixture model (GMM) to approximate the distribution of FrFT modular detector. Theoretical analysis and simulation results show that it can retain the high estimation accuracy and also greatly reduce the computational complexity at the same time.

Calculation of Discrete Fractional Fourier Transform based on Adaptive LMS algorithm

This paper establishes a relation between the least mean square (LMS) algorithm and the discrete fractional Fourier transform (DFRFT). It is shown that the LMS algorithm could provide a new way for the calculation of DFRFT by properly choosing the input vector and adaptation speed