This thesis presents techniques to detect mental blocks in humans based on the physiological parameters skin potential and skin resistance. We examine physiological measures from the Musico Cause and Effect study of the Science Network for Man and Music at the University of Music and Dramatic Arts, Mozarteum Salzburg. The existing digital signal analysis tool AIDA used to process the physiological data has been replaced by the Dynalyzer developed by the author with considerable improvements in accuracy and performance. We present fundamentals of digital signal processing, outline the measurement of physiological data, and discuss characteristics of mental blocks. We suggest several criteria for the detection of mental blocks based on characteristic features of the physiological time series in the time and/or frequency domain. In first experiments the potential of these criteria is evaluated by applying them to actual physiological data of a test subject. As no ground truth on the occurrence of mental blocks is available, the experimental results can only be an indicator for the quality of the detection of mental blocks. Further experiments are conducted with data from the Vienna determination test assessing the reactive stress tolerance and attention deficits of human test subjects.
Table of Contents
Affidavit
Acknowledgement
Details
Keywords
Abstract
List of Figures
List of Tables
1 Introduction
1.1 Mental Blocks
1.2 Thesis Structure
2 Digital Signal Analysis
2.1 Analog to Digital Conversion
2.2 The Fourier Transformation
2.2.1 The Fast Fourier Transformation . .
2.2.2 The Discrete Cosine Transformation
2.3 Windowing
2.3.1 Window Size
2.3.2 Window Functions
2.3.3 The Short Time Fourier Transformation
2.4 Signal Analysis with Correlation Methods
2.5 The Power Density Spectrum
2.6 Digital Filtering
2.6.1 The Moving Average Filter
2.6.2 The Windowed-Sinc Filter
3 Electrodermal Activity
3.1 Measurement
3.1.1 Examined Parameters
3.1.2 Measurement Problems
3.2 Analysis of Electrodermal Activity
4 Data Analysis
4.1 Physiological Time Series
4.2 The Analysis Tool AIDA
4.2.1 Filtering and Windowing
4.2.2 Detection of the Main Component .
4.2.3 Calculating the Model Function .
4.2.4 Shortcomings of AIDA
4.3 The Analysis Tool Dynalyzer
4.3.1 Filtering and Windowing
4.3.2 Detection of the Main Component .
4.3.3 Improvements in the New System
4.3.4 The Graphical User Interface . .
4.4 Testing of the Analysis Tools
4.4.1 Single Sinusoids
4.4.2 Multiple Sinusoids
4.4.3 Frequency Responses
5 Mental Blocks
5.1 Mental Block Situations
5.2 Characteristics of Mental Blocks
5.2.1 Periodic System of Regulatory States
5.2.2 Dynamic Function
5.3 Detection Criteria
5.3.1 High Activity Time
5.3.2 High Power
5.3.3 Low Power
5.3.4 Low to High Ratio
5.3.5 Total Power
6 Experimental Results
6.1 A Comparison of Mental Block Criteria
6.2 The Vienna Determination Test
7 Conclusion
7.1 Outlook
7.2 User Interface Optimization based on Physiological Measures
Bibliography
List of Abbreviations
Appendix
Index
Acknowledgement
The Book of Nature is written in mathematical characters.
Galileo Galilei, Italian scientist (1564-1642)
First of all I want to thank my parents for all the support during my studies.
I want to thank Ao. Univ. Prof. Dipl-Ing. Dr. Helmut A. Mayer, for the continuous and inexhaustible support during the writing of this diploma thesis and the development of the software.
Further I want to thank Prof. Dr. Hans-Ullrich Balzer, who had the idea of this diploma thesis and gave me so much important information and useful advices during the implementation and the writing of this thesis.
Furthermore I want to thank all people from the ”Science Network for Man and Music” of the ”University of Music and Dramatic Arts”, Salzburg, who supported me with information about their research especially Dr. Florentina M. Fritz and Dr. Kai Bachmann.
Additionally I want to thank all of those, who helped me during the writing of my diploma thesis, especially DI (FH) Michael Hofer for the attentive reading of my work.
Details
illustration not visible in this excerpt
Keywords
illustration not visible in this excerpt
Abstract
This thesis presents techniques to detect mental blocks in humans based on the physi- ological parameters skin potential and skin resistance. We examine physiological mea- sures from the Musico Cause and Effect study of the science network for man and music of the University of Music and Dramatic Arts, Mozarteum Salzburg. The exist- ing analysis tool of the science network, AIDA, is the basis for our analysis tool, the Dynalyzer, which was developed by the author. A comparison of both systems shows considerable improvements in accuracy and performance of our system. We outline the measurement methodology of physiological parameters and present fundamentals of signal processing. Characteristics of mental blocks are discussed in physiological and neuronal context. We suggest several criteria for the detection of mental blocks based on characteristic features of mental blocks in the frequency or the time domain. First experiments are conducted using these criteria. As for our physiological data no ground truth on the occurrence of mental blocks is available, the experimental results can only be an indicator for the quality of the detection of mental blocks. Further experiments are conducted with data from the Vienna determination test assessing the reactive stress tolerance and attention deficits of human test subjects.
List of Figures
2.1 The process of digitizing
2.2 The decomposition of the Fast Fourier Transformation. After31.
2.3 The correlation of two signals with sinusoids. After31
2.4 The rectangular window
2.5 The Bartlett window
2.6 The Hanning or cosine Bell window
2.7 The Hamming window
2.8 The Gaussian window function
2.9 The Blackman window
2.10 Spectral leakage from a rectangular window
2.11 Properties of the step response
2.12 Properties of the frequency response
2.13 The frequency response of the moving average filter
2.14 The frequency response of a perfect low-pass filter
2.15 The Sinc-function in a discrete interval of time
2.16 The Windowed-Sinc filter kernel with the Blackman window.
2.17 The frequency response of the Windowed-Sinc filter
3.1 The SMARD-Watch10
4.1 Example of two model functions found in a window with 60 samples ([10]by permission of the author)
4.2 The Dynalyzer, Version 1.4
4.3 The power density spectrum of a sinusoid calculated with the Dynalyzer.
4.4 The power spectrum of a sinusoid calculated with AIDA
4.5 Frequency response of the Dynalyzer
4.6 Frequency response of AIDA
5.1 The periodic system of regulation states ([10, 3] by permission of the author)
5.2 The analysis of skin potential and skin resistance with AIDA
6.1 Mental Blocks detected with different criteria with fixed windows
6.2 Mental Blocks detected with different criteria with dynamic windows. .
6.3 Detected mental blocks compared to the score function of the Vienna determination test
List of Tables
2.1 Four types of Fourier Transformations
2.2 Properties of several common window functions35
4.1 The test configurations
4.2 The detection of sinusoids with the Dynalyzer and with AIDA
4.3 The detection of multiple sinusoids with the Dynalyzer and with AIDA.
5.1 Losses in the meritocracy under stress
6.1 Settings of mental block criteria
6.2 Mental block detection results using different criteria
6.3 The score function of Vienna determination test reactions
6.4 Reactions during detected mental blocks
Introduction
The idea for this thesis originated in the search for an appropriate way to evaluate the cognitive and emotional effects of user interfaces on humans in real time. The evaluation of physiological measures may help us to describe the current mental state of humans 3. Therefore, we decided to investigate those measures.
In todays biofeedback research the change of physiological parameters like skin resistance or skin potential after a stimulus are examined. For example, in clinical applied biofeedback, subjects receive on-line feedback of those parameters and are trained to increase or decrease them above or below a baseline level depending on the presented stimuli. The purpose is to self-control body regulation, which is important in the struggle against a disease like epilepsy5.
Besides the recovery from diseases the evaluation of physiological measures is used for several other purposes. In a study about Musico Cause and Effect 3of the science network for man and music at the University of Music and Dramatic Arts, Mozarteum Salzburg, physiological data were analyzed and correlated with characteristic features of music. Particularly, the parameters volume change, change of frequency, and tone interval density of music were examined, while a listeners physiological values were monitored. The influence of music on concentration and relaxation was examined using a psychological test3.
In the study a relation between music and physiological data and the influence of music on psychological and behavioral parameters like concentration and endurance was found3. Therefore, using these techniques the cognitive and emotional effects of user interfaces on humans could be evaluated by analyzing physiological data. However, here we restrict ourselves to the analysis of features in physiological data.
1.1 Mental Blocks
In a study9of the science network for man and music the psychological and the physiological reactions to stress of performing musicians were investigated. It was found that nervosity and stress may lead to a Mental Block (MB) of an artist9. During the investigations of the science network characteristic features in the physiological data were observed, while MBs occurred [9, 3].
Particularly, these features were studied after a cognitive or emotional stress situation. The cognitive condition of a test person was examined by the analysis of the parameter skin potential and the emotional condition by the parameter skin resistance. The physiological signals are searched for periodic features in certain intervals of time. The periodic component with the maximal power density in that interval is defined to be the Main Component (MC). It is considered to be characteristic for the cognitive or emotional state of the test subject in the examined interval of time [3, 10].
The MCs in the physiological data are analyzed with the purpose to detect MBs. Based on the results of the Musico Cause and Effect study, an MB is defined by a sudden change in frequency of the MC from high to very low values. The research of Balzer3suggests that a high frequency of the MC is characteristic for a high cognitive, or emotional demand, and a low frequency for relaxation. MBs, characterized by an MC with a low frequency, are the sign of relaxation of regulatory processes in the nervous system after a cognitive or emotional overload [3, 10].
The detection of MBs in this thesis is based on the analysis of periodicities in the physiological measures skin potential and skin resistance of the Musico Cause and Effect study 3. The existing analysis tool AIDA, which was developed by Balzer 3, is the basis for our analysis tool, the Dynalyzer, which was developed by the author. The new system includes the analysis techniques of AIDA and other signal analysis techniques. The Dynalyzer was implemented because AIDA has several limitations concerning the detection of MBs. Besides the higher accuracy, the Dynalyzer is also faster based on the usage of the Fast Fourier Transform (FFT) for the analysis. We compared the performance of both systems by their frequency responses and by detecting the MC in a signal containing single and multiple sinusoids with different amplitude.
We suggest several criteria to detect MBs, which identify the characteristic feature of an MB either in the frequency or in the time domain. The MBs detected with the different criteria are compared by the start of their occurrence, their duration, and their total number.
MBs can arise from different stress situations mostly under high cognitive or emotional encumbrance and time pressure. The detection of MBs with a combination of two suggested criteria is evaluated in a psychological test, the Vienna Determination Test (VDT), assessing the reactive stress tolerance, attention deficits, and reaction speed. We examine the co-occurrence of the detected mental blocks with faulty reactions during the VDT. The VDT is just a first step to evaluate the ability to detect MBs because it does not provide ground truth about the occurrence of MBs .
1.2 Thesis Structure
The introduction (section 1) describes the motivation and the background of the investigations on physiological measures, shortly characterizes Mental Blocks (MB) and outlines the content of this thesis.
In section 2 we present fundamentals of signal processing. Among others, the windowing of signals, methods to calculate the Power Density Spectrum (PDS), and two digital filters are discussed.
Details about the examined physiological parameters, the device used to measure these parameters, and the preprocessing of these time series data are presented in section 3.
The methodology of the analysis of physiological data with AIDA and with the Dyna- lyzer is described in section 4. It discusses the problems in performance of AIDA and shows the improvements of the analysis with the Dynalyzer in a comparison of both systems.
In section 5 mental blocks are described in several examples, and their characteristics are discussed in physiological and psychological context. The results of the investi- gations on MBs of Balzer3are presented, and we suggest several criteria to detect MBs.
The results of the experiments with the MB criteria are presented in section 6. A comparison of MBs detected using different criteria is discussed and the detection using two combined criteria is evaluated with the Vienna Determination Test (VDT).
In section 7 we summarize the results of the detection of MBs and present an outlook to possible improvements and applications of the proposed techniques.
Digital Signal Analysis
The analysis of physiological signals requires several signal analysis fundamentals and mathematical definitions and methods, which will be discussed in the following sections.
The term signal analysis generally refers to the science of analyzing or interpreting signals that can be transient in nature, occurring only over a short time duration or periodic, or they may be random and unpredictable. Methods of signal analysis apply to all of these types of signals32.
2.1 Analog to Digital Conversion
Before a physical quantity can be analyzed in a computer, it has to be measured and digitized. These processes shown in figure 2.1 are discussed in the following paragraphs. When a physical quantity is measured by a transducer it could contain higher frequencies than those we want to measure. To avoid aliasing effects the highest frequency in the signal should be below the Nyquist-frequency, which will be discussed in the next paragraph. The analog anti-alias filter removes these frequencies before the voltage is quantized by the analog-to-digital converter. The sample-and-hold unit assures that during quantization the measured voltage is kept constant31.
The sequence of numbers or sample sequence produced by the analog-to-digital con- verter is fetched by the sample-and-hold unit in intervals of time known as sampling time. The inverse of the sampling time refers to the sampling frequency. According
Figure 2.1: The process of digitizing.
illustration not visible in this excerpt
to the sampling theorem of Nyquist and Shannon the sampling frequency has to be greater than twice the Nyquist-frequency in order to represent the signal exactly by samples. The Nyquist-frequency is the highest frequency in the analog signal after the anti-alias filtering [31, 32].
The output signal in figure 2.1 can now be analyzed with methods presented in the following sections.
2.2 The Fourier Transformation
The Fourier analysis is named after Jean Baptiste Joseph Fourier (1768-1830), who first had the idea to decompose signals into sinusoids. Nowadays, four variants of Fourier Transformation (FT)s exist: The Continuous Fourier Transform (CFT), the Fourier Series, the Discrete Fourier Transform (DFT), and the Discrete Time Fourier Transform (DTFT). The Fast Fourier Transform (FFT) is an algorithm to calculate the DFT in an efficient way31.
The purpose of FTs is to transform a function from the time to the frequency domain and vice versa. A discrete signal can be represented mathematically as a continuous function multiplied by an impulse train. This is a periodic sequence of unit impulses. The unit impulse or Kronecker delta function is defined as31:
(2.1)
illustration not visible in this excerpt
A signal in the time domain can be either continuous or discrete, and it can be periodic, or aperiodic. This translates to the four types of FTs31:
illustration not visible in this excerpt
Table 2.1: Four types of Fourier Transformations.
If a signal is continuous, and aperiodic or periodic in the time domain, it will be also aperiodic in the frequency domain, when transformed with a continuous FT. If a signal is discrete in the time domain it will be periodic in the frequency domain when transformed with a discrete FT. Each of these four transforms has a general complex version and a real version31. The general CFT is defined as31:
(2.2)
illustration not visible in this excerpt
[...]
-
-
-
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen.