Weibei Dou

A framework of fuzzy information fusion for the segmentation of brain tumor tissues on MR images

A framework of fuzzy information fusion is proposed in this paper to automatically segment tumor areas of human brain from multispectral magnetic resonance imaging (MRI) such as T1-weighted, T2-weighted and proton density (PD) images. A priori knowledge about tumors described by radiology experts for different types of MRI are very helpful to guide a automatic and a precise segmentation. However, the terminology used by radiology experts are variable in term of image signal. In order to benefit of these descriptions, we propose to modellize them by fuzzy models. One fuzzy model is built for one type of MRI sequence. The segmentation is finally based on a fusion of different fuzzy information obtained from different types of MRI images. Our algorithm consists of four stages: the registration of multispectral MR images, the creation of fuzzy models describing the characteristics of tumor, the fusion based on fuzzy fusion operators and the adjustment by fuzzy region growing based on fuzzy connecting. The comparison between the obtained results and the hand-tracings of a radiology expert shows that the proposed algorithm is efficient. An average probability of correct detection 96% and an average probability of false detection 5% are obtained through studies of four patients.

Features extraction and selection for emotional speech classification

The classification of emotional speech is a topic in speech recognition with more and more interest, and it has giant prospect in applications in a wide variety of fields. It is an important preparation for automatic classification and recognition of emotions to select a proper feature set as a description to the emotional speech, and to find a proper definition to the emotions in speech. The speech samples used in this paper come from Berlin database which contains 7 kinds of emotions, with 207 speech samples of male voice and 287 speech samples of female voice. A feature set of 50 potentially features is extracted and analyzed, and the best features are selected. A definition of emotions as 3-states emotions is also proposed in this paper.

Recognition of emotions in speech by a hierarchical approach

This paper deals with speech emotion analysis within the context of increasing awareness of the wide application potential of affective computing. Unlike most works in the literature which mainly rely on classical frequency and energy based features along with a single global classifier for emotion recognition, we propose in this paper some new harmonic and Zipf based features for better speech emotion characterization in the valence dimension and a multistage classification scheme driven by a dimensional emotion model for better emotional class discrimination. Experimented on the Berlin dataset with 68 features and six emotion states, our approach shows its effectiveness, displaying a 68.60% classification rate and reaching a 71.52% classification rate when a gender classification is first applied. Using the DES dataset with five emotion states, our approach achieves an 81% recognition rate when the best performance in the literature to our knowledge is 76.15% on the same dataset.

Multi-stage classification of emotional speech motivated by a dimensional emotion model

This paper deals with speech emotion analysis within the context of increasing awareness of the wide application potential of affective computing. Unlike most works in the literature which mainly rely on classical frequency and energy based features along with a single global classifier for emotion recognition, we propose in this paper some new harmonic and Zipf based features for better speech emotion characterization in the valence dimension and a multi-stage classification scheme driven by a dimensional emotion model for better emotional class discrimination. Experimented on the Berlin dataset with 68 features and six emotion states, our approach shows its effectiveness, displaying a 68.60% classification rate and reaching a 71.52% classification rate when a gender classification is first applied. Using the DES dataset with five emotion states, our approach achieves an 81% recognition rate when the best performance in the literature to our knowledge is 76.15% on the same dataset.

What is the best segment duration for music mood analysis

Music mood present the inherent emotional state of music on certain duration of music segment. However, the mood may vary in the music pieces. Thus it is important to investigate the duration of music segments which can best present the stable mood states in music. Four versions of music datasets with duration of clips from 4 seconds to 32 seconds are tested in this paper, and a cross test between these four version are also experimented. Better classification rates are obtained with the music segments of 8 seconds and 16 seconds. The segments with duration of 32 seconds are proved as might be not stable concerning the music mood states.

Two-stage Classification of Emotional Speech

The purpose of this paper is to make an automatic classification of speech into seven emotional classes as anger, boredom, disgust, fear, gladness, neutral and sadness. A two-stage classification composed of several sub-classifiers is proposed. A feature set with 68 features has been computed over 286 speech samples from the Berlin database. The sequential forward selection method (SFS) has been used for each classifiers of the two stages in order to decide the feature subsets in each step. The result for the first stage as three-state classification is 87%, and the global result of the seven emotional classes is 78%, where the correct recognition rate of random classification by chance is about 15%

MDCT Sinusoidal Analysis for Audio Signals Analysis and Processing

The Modified Discrete Cosine Transform (MDCT) is widely used in audio signals compression, but mostly limited to representing audio signals. This is because the MDCT is a real transform: Phase information is missing and spectral power varies frame to frame even for pure sine waves. We have a key observation concerning the structure of the MDCT spectrum of a sine wave: Across frames, the complete spectrum changes substantially, but if separated into even and odd subspectra, neither changes except scaling. Inspired by this observation, we find that the MDCT spectrum of a sine wave can be represented as an envelope factor times a phase-modulation factor. The first one is shift-invariant and depends only on the sine waves amplitude and frequency, thus stays constant over frames. The second one has the form of sinθ for all odd bins and cosθ for all even bins, leading to subspectras constant shapes. But this θ depends on the start point of a transform frame, therefore, changes at each new frame, and then changes the whole spectrum. We apply this formulation of the MDCT spectral structure to frequency estimation in the MDCT domain, both for pure sine waves and sine waves with noises. Compared to existing methods, ours are more accurate and more general (not limited to the sine window). We also apply the spectral structure to stereo coding. A pure tone or tone-dominant stereo signal may have very different left and right MDCT spectra, but their subspectra have similar shapes. One ratio for even bins and one ratio for odd bins will be enough to reconstruct the right from the left, saving half bitrate. This scheme is simple and at the same time more efficient than the traditional Intensity Stereo (IS).

Facial expression recognition and generation using sparse autoencoder

Facial expression recognition has important practical applications. In this paper, we propose a method based on the combination of optical flow and a deep neural network - stacked sparse autoencoder (SAE). This method classifies facial expressions into six categories (i.e. happiness, sadness, anger, fear, disgust and surprise). In order to extract the representation of facial expressions, we choose the optical flow method because it could analyze video image sequences effectively and reduce the influence of personal appearance difference on facial expression recognition. Then, we train the stacked SAE with the optical flow field as the input to extract high-level features. To achieve classification, we apply a softmax classifier on the top layer of the stacked SAE. This method is applied to the Extended Cohn-Kanade Dataset (CK+). The expression classification result shows that the SAE performances the classification effectively and successfully. Further experiments (transformation and purification) are carried out to illustrate the application of the feature extraction and input reconstruction ability of SAE.

Histogram-Based generation method of membership function for extracting features of brain tissues on MRI images

We propose a generation method of membership function for extracting features of brain tissues on images of Magnetic Resonance Imaging (MRI). This method is derived from histogram analysis to create a membership function. According to a priori knowledge given by the neuro-radiologist, such as the features of gray level of differentiate brain tissues in MR images, we detect the peak or valley features of the histogram of MRI brain images. Then we determine a transformation of the histogram by selecting the feature values to generate a fuzzy membership function that corresponds to one type of brain tissues. A function approximations process is used to build a continuous membership function. This proposed method is validated for extracting whiter matter (WM), gray matter (GM), cerebra spino fluid (CSF). It is evaluated also using simulated MR images with two different, T1-weighted, T2-weighted MRI sequences. The higher agreement with the reference fuzzy model has been discovered by kappa statistic.

Dual-mode switching used for unified speech and audio codec

This paper presents a dual-mode switching method between time-domain codec and transform-domain codec of audio coding. It is a key technique of unified speech and audio (music) coding, since the replaying audio quality corresponds to the suitable codec selection and smooth switching between them. The proposed method consists of two steps, codec mode selection and switching. The binary decision trees (BDTs) algorithm is used to take a decision of mode selection, because of its advantages of high accuracy, low delay and low complexity. For smoothing transition between two codec, a pre-coding strategy is suggested in this paper. The classical speech codec, Algebraic Code Excited Linear Prediction (ACELP) and the Advanced Audio Coding (AAC) of MPEG are used for validating the proposed method. The scores of PESQ with 11 testing sequences show that the proposed switching method will not bring additional noise and can get higher objective evaluation of audio quality than single codec.