Dissertation > Excellent graduate degree dissertation topics show

A Research of Emotion Recognition Based on Respiratory Signals

Author: WangBin
Tutor: LiuGuangYuan
School: Southwestern University
Course: Signal and Information Processing
Keywords: Emotion Recognition Respiratory Signal Wavelet Transform Genetic Simulated Annealing Algorithm
CLC: TP391.4
Type: Master's thesis
Year: 2010
Downloads: 184
Quote: 2
Read: Download Dissertation

Abstract


Affective computing is to make computer own, understand, and moreover, recognize human emotions. Emotion recognition is an important area in affective computing and it is the basis of building a harmonious man-machine environment. The research contents of emotion recognition include facial expressions, speech sound, body postures, and physiological signals and so on. The study of physiological signals is the most difficult. And Respiratory (RSP) signal is the main object in the emotion recognition based on physiological signal. The change of RSP signal is one of the most important and the realest embodiment of human emotions, so that through the study of the RSP signal, we can recognize people’s inner feelings and emotional changes.As the emotional RSP signal contains a wealth of emotional information, and can obviously reflect the changes of human emotional state. Therefore, in this paper, features were extracted from the RSP signals. And intelligent optimization algorithms were used to select RSP emotional features. To a certain extent, the algorithm effectively deals with the limitations of feature selection by traditional methods. On this basis, the paper makes further research on RSP-based feature combinations which can represent specific emotions. The process of emotion recognition based on RSP signals includes four steps:RSP signal acquisition, feature extraction, feature selection, and classification recognition.Collecting valid RSP signals is the first step in emotion recognition. RSP data were collected in the state of joy, surprise, disgust, grief, angry and fear by the acquisition equipment MP150, which creates an emotional RSP signal database. The subjects were freshmen from Southwest University. And movie clips rich in emotion were used as the inspired material.Extracting effective features from RSP signal is essential to research on emotion recognition. Before feature extraction, the collected RSP data were firstly preprocessed:Butterworth low-pass filter is used for filter processing, and then the data is normalized by the baseline value. Wavelet Transform has good time-frequency characteristics, thus as a telling time-frequency analysis method, it can effectively extract minutiae feature for multi-resolution analysis. In this paper,84 wavelet coefficient features were extracted from the collected RSP signals. And through analysis of its own characteristics about the RSP signal,87 statistical features were extracted. And finally the 171-dimensional original feature set was formed, which was used to recognize emotion.Feature selection is a combinatorial optimization problem. So we can solve the feature selection problem through the way of solving combinatorial optimization problems. And intelligent optimization algorithms are a good solution for combinatorial optimization problems. Genetic Algorithm (GA) is a method of simulating natural evolutionary processes to search the optimal solution. As the overall search strategy and optimization searching method of GA only need to know the objective function of searching direction and the corresponding fitness function, so GA provides a general framework for solving the complex system problems, and it is widely used in the function optimization, combinatorial optimization, production scheduling problems, automation, robotics, image processing, artificial life, and genetic coding. Simulated Annealing algorithm (SA), instead of relying on the initial conditions, moves up and down in the search space, it specializes in solving multi-dimensional problem, and it can deal with any degree of non-linear, discontinuous and random problems. SA has been successfully applied in combinatorial optimization, neural networks, and image processing and code design. In a word, GA performs well in global search, but badly in local search; and SA has a strong ability of local search. So the two algorithms can learn from each other, and Genetic Simulated Annealing Algorithm (GASA) is generated through combining SA with GA. On the basis of introducing the dimension reduction strategy, GASA was combined with Fisher classifier. And the correct recognition rate of Fisher classifier was used as the evaluation criteria function in GASA. The six emotions were recognized and effective feature combination respectively representing specific emotional states were selected.Through the simulation experiments of the affective data, the one-to-one and one-to-many recognition methods were adopted in this paper. We verified the validity of GASA combined with the Fisher classifier for the RSP signal emotion recognition. Thus, not only good emotion recognition rates, but also effective feature subsets for emotion recognition were obtained.

Related Dissertations

  1. Feature Extraction, Selection and Combination in Lipreading,TP391.41
  2. Research on Identification System of Cashmere and Wool Fiber,TS101.921
  3. Research on Feature Selection and Construction in Emotion Speech Recognition,TP18
  4. Characteristics of sensory stimulation evoked,R318.0
  5. Network transmission ROI image coding algorithm,TN919.81
  6. Based on Contourlet Transform Digital Watermarking Method,TP309.7
  7. Image Fusion Algorithms Based on Multi-scale Analysis,TP391.41
  8. Feature Extraction Technologies Research and Implementation of 3D Models Based on Wavelet Transform,TP391.41
  9. An Algorithm on Clustering and Anomaly Detection for Multiple Data Streams,TP311.13
  10. Research on Information Hiding Based on Images,TP309.7
  11. SOPC Design of Transient Power Quality Disturbances Detecting Based on Nios Ⅱ,TN47
  12. Based on Lifting Scheme Wavelet Packet Transform and Artificial Neural Network Wing-box Multi-damage Detection,V224
  13. An Empirical Study on the Relationship between the China Stock Index Futures and Spot Markets,F224
  14. A Study Regarding to Organizational Identification and Simulation of Reality for Double-Organization Fabric,TS105
  15. Based on the difference between the pixel grayscale image denoising wavelet edge detection and combining adjacent to its,TP391.41
  16. Combined DWT Dynamic Data Reconciliation Research and Application,TP274
  17. Fog image enhancement method,TP391.41
  18. The Compression and Application of the Brain CT Image,TP391.41
  19. Two Algorithms for Image Processing Based on Chaos and Fractal,TP391.41
  20. Reasearch on Packing of Apparel Shaped Parts Using Genetic Simulated Annealing Algorithm,TP301.6
  21. Helicopter main gearbox fault diagnosis and failure prediction technology research,V267

CLC: > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Pattern Recognition and devices
© 2012 www.DissertationTopic.Net  Mobile