Printer Friendly

An attempt to evaluate mental workload using wavelet transform of EEG.

INTRODUCTION

In human-computer interaction, many tasks are performed using complicated cognitive information processing. Such situations cause mental workload or stress that has not been observed as a result of performing traditional physical tasks. As a result of workers' reduced agility caused by excessive workload, productivity decreases and human error frequently occur. In particular, mental workload induced by overloading the capacity of working memory would be a limiting factor in the early stage of acquiring computer procedural skills. An effective method of monitoring mental workload would therefore be useful in the evaluation of alternative interface designs.

Heart rate variability (HRV) is frequently used as a measure for evaluating mental workload (Aasman & Mulder, 1987; Akselrod et al., 1981; Bartoli, Baselli, & Cerutti, 1985; Baselli & Cerutti, 1985; Baselli et al., 1987; Cerutti, Fortis, Libeart, Baselli, & Civardi, 1988; Luczak & Laurig, 1973; Murata, 1994; Tsuji & Mori, 1994). Cerutti et al. (1988) applied an autoregressive model to analyze the fluctuation of R-R intervals and showed that the difference in mental activity between a resting condition and that required to perform a mental arithmetic task could be evaluated on the basis of a low-frequency component at around 0.1 cycles/beat and a high-frequency component corresponding to the respiration-frequency band. Murata (1994) also indicated that mental workload caused by demanding tasks such as arithmetic could be evaluated using an autoregressive power spectrum if the respiration interval was carefully controlled. HRV is a measure that indicates the activity of the autonomic nervous system. Physiological measures such as an electroencephalogram (EEG) might also be promising, especially for the evaluation of mental workload induced by activities such as memory tasks.

EEG measures have been shown to be highly sensitive to variations in task difficulty (Gevins et al., 1998; Gevins, Smith, McEvoy, & Yu, 1997; Gevins, Zeitlin, Doyle, Schafer, & Callaway, 1979; Gevins, Zeitlin, Doyle, Yingling, et al., 1979; Gundel & Wilson, 1992; Humphrey & Kramer, 1994; Wilson & Fisher, 1995). In these traditional analyses, the power spectral density of EEG signals is calculated using fast Fourier transform (FFT) to examine the change of frequency characteristics. Such an approach allows one to understand how the ratio of a specific frequency band such as an alpha band changes with the accumulation of mental fatigue (Okogbaa, Shell, & Filipusic, 1994) or when the mental work level changes (Gevins et al., 1998). The power spectrum represents frequency characteristics during a predetermined interval. If EEG is measured at different time epochs, one can roughly know how the frequency characteristics change for different time epochs. Unlike time series data, however, the time characteristics during the interval cannot be known.

It has been shown that the P300 component of the event-related potential (ERP) can be an effective measure of mental or cognitive workload (Johnson & Donchin, 1980; Kramer, Wickens, & Donchin, 1985; Ullsperger, Metz, & Gille, 1988). The P300 amplitude increases with increased cognitive work intensity. However, it takes a great deal of time to obtain an ERP waveform 2based on each EEG recording. Cognitive information processing is conducted in the central nervous system in a very short period. Such a time-consuming measurement using ERP (P300) has the potential to miss the change in cognitive information processing according to the difference of mental work loading.

EEG signals are, usually, nonstationary and change abruptly in a short period, which leads to the change of frequency characteristics. It is impossible to detect such a change of frequency characteristics by means of traditional approaches using power spectral analysis of EEG or the P300 component of ERR Theoretically, signal processing of EEG signals by FFT assumes that the signals are stationary. Biological signals are composed of phenomena that change every moment, often abruptly. A time series that does not contain such changes can be regarded as stationary. Actually, biological signals such as EEG and R-R intervals are not stationary. The wavelet analysis or short-time Fourier transform (STFT), which can investigate the time-frequency characteristics of biological signals, is expected to extract more useful information that cannot be extracted with traditional approaches.

The STFT is a useful technique in that it allows one to depict nonstationary signals as stationary ones by using a window function. The time-frequency resolution of STFT is fixed over an entire time-frequency plane once a window is chosen. In such nonstationary signals, the change in stress or workload factors results in an abrupt change in biological signals. The STFT does not properly treat such changes in biological signals. Better resolution in time at higher frequencies (rapid changes in time) is needed. At low frequencies, better resolution in frequencies is required. To overcome these problems, the wavelet transform (Daubechies, 1988, 1990) is promising because it changes the window size adaptively and uses short windows at high frequencies and long windows at low frequencies. Consequently, it is expected that the wavelet transform is applicable to nonstationary signals.

In spite of this, biological signals such as R-R intervals and EEG are analyzed using an arbitrary analysis interval that is assumed to be stationary. It is reasonable to assume that the mental state or mental workload is nonstationary and changes every moment. Global analysis using a traditional signal processing technique such as FFT and local analysis that investigates the time-frequency characteristics of biological signals are both necessary in the evaluation of mental workload. Therefore, it is promising to track how a frequency component of EEG signals changes with time so that mental workload can be evaluated more accurately and confidently.

The aim of this study was to evaluate mental workload induced during human-computer interactions using wavelet transform of EEG signals. The three levels of task difficulty were prepared in a continuous matching task, and EEG signals were measured continuously during the experimental task. The parameters extracted from the wavelet transform of EEG signals were used to evaluate mental workload. The correspondence between these parameters and the performance measures or the psychological rating of mental workload was investigated in order to propose an effective measure of mental workload.

METHODS

Participants

Eight healthy male undergraduate and graduate students 21 to 24 years old participated in the experiment.

Apparatus

Using a polygraph system (NEC Sanei, SYN-AFIT2200), EEG signals were recorded during a task at each work level. The EEG signals were band-pass filtered at 0.03 to 60 Hz and sampled at 1 kHz using an A/D converter (A/D Instruments, MacLab/8).

Experimental Task

The experimental task was a continuous matching task that required the participant to indicate whether a current stimulus matched a stimulus presented on a previous trial, according to Gevins et al. (1998). The stimuli, presented on a 19-inch (48-cm) color monitor (Mitsubishi, FHL61155-SATK]) located in front of the participant, were 12 single capital letters (A-L) presented for 200 ms every 4.5 s. The presentation of stimuli was controlled using a color audiovisual tachistoscope (Iwatsu, IS-701D) connected to a personal computer (Apple, Power Macintosh 8100/80AV). The possible stimulus locations were randomly selected at eight points on the CRT. The identity of the letter and its spatial position varied randomly from trial to trial.

The participant performed the tasks at each of three levels of difficulty. In the low work level, the participants were required to compare the identity and location of the current stimulus with that of a stimulus presented on the previous trial. When the identity and location of the current stimulus matched with that of the previous stimulus, the participant was asked to press the 1 key. When the stimuli did not match, the participant was required to press the 2 key. In the moderate load level, the participant compared the current stimulus with that presented two trials ago. The high load level required the participant to compare the current stimulus with that presented three trials ago. A matching stimulus occurred randomly on 50% of the trials. The participants were required to remember the identity and location of each stimulus as well as its sequential order for 8 and 13.5 s in the moderate and high load levels, respectively.

Procedure

Task difficulty was a within-subject variable. In one session, the participant was required to perform 50 trials for each of the three task difficulties. The participants performed two sessions on the same day. The order of the three task difficulty conditions within each session was randomized to eliminate any order effects. The experimental task was different from that of Gevins et al. (1998) in that the participant had to remember both the identity and location of each stimulus. The time interval between task difficulty conditions, each consisting of 50 trials, was about 1 min. The participants were also permitted to rest between sessions for about 5 min. The participants were instructed to respond as quickly and as accurately as possible.

EEG signals during the task were continuously recorded from Fz, Cz, and Pz according to the international Jasper 10-20 method. Electrooculogram activity was also recorded from electrodes located above each eye that were referenced to an electrode at the outer canthus of each eye. Performance (reaction time and errors trials, if any) was also recorded continuously. The EEG data were visually inspected, and data segments containing eye movements and blinks, instrumental noise, and movement artifacts were eliminated from subsequent analyses.

The control of respiration frequency, as already pointed out, is necessary for the proper evaluation of mental workload using HRV (Murata, 1994). At present, the general effect of respiration on EEG is not well known. The respiration was controlled at an interval between 3 and 4 s for the sake of caution and to reduce movement artifact in the EEG record. Respiration was controlled as follows. The participants learned to breathe with moderate depth and according to a respiration interval between 3 and 4 s using a metronome. The learning period differed among participants and ranged from 5 to 10 min. When the experimenter judged that the participant could breathe with moderate depth and according to the predetermined interval, the experimental session was started. During the experiment, respiration was recorded to monitor the respiration interval and depth. When the participant did not breathe at moderate depth at the 3- to 4-s interval, the measurement was started again.

For each of the three levels of difficulty, the participant was required to evaluate mental workload after the performance of each task using the NASA-Task Load Index (NASA-TLX; Hart & Staveland, 1988; Hill et al., 1992). Only the scoring phase was conducted; the workload parameter evaluation (paired comparison) phase of the NASA-TLX was not administered because it would have required many participants and more time for data reduction and analysis. The omission of the workload parameter evaluation phase was based on the finding of Hill et al. (1992) that the paired-comparison sort procedure may be skipped without compromising the workload measure. The items were overall workload, mental demand, physical demand, temporal demand, performance, effort, and frustration level.

Analysis

The wavelet transform is a time-frequency analysis method that maps the signal x(t) (one-dimensional time domain) into a two-dimensional function on a time-frequency plane by moving the window function along the time axis and changing the size of the window function according to frequencies (Coifman & Wickerhauser, 1993; Meyer & Ryan, 1993; Young, 1993). This is analogous to a musical score. The wavelet transform of the signal x(t) is given by

(1) W(a, b) = [[integral].sup.[infinity].sub.-[infinity]] x(t)[phi](t - a/b)dt,

in which a and b are scale (dilation) and shift (delay) parameters, respectively. The term x(t) corresponds to the time series signal. The function [phi] is referred to as an analyzing (basic) wavelet. The parameters a and b correspond to the time and frequency axes, respectively. The wavelet transform can adaptively vary the time-frequency resolution by using short windows at high frequencies and long windows at low frequencies. The Gabour function (Daubechies, 1988, 1990) was used as a analyzing (basic) wavelet, because it can relatively localize the width of the time-frequency window. The Gabour function, [phi], is given by

(2) [phi](t) = exp(-[(t - b)/a].sup.2])exp(-j[[omega].sub.0]t - b/a),

in which [[omega].sub.0] is an empirically determined constant (in this study, [[omega].sub.0] = 2[pi]). The discrete wavelet transform is given by

(3) W(m[DELTA]F, b[DELTA]t) = [n.summation over k=1]x(k[DELTA]t) exp(-(m[DELTA]f[(k-b)[DELTA]t).sup.2]) x exp(-j2[pi]m[DELTA]f(k-b)[DELTA]f),

k - 1, 2, ..., n; m = 1, 2, ..., M,

in which [DELTA]t, n, M, and [DELTA]f represent a sampling interval, number of data points, maximum number of octave division, and basic frequency, respectively. The term x([DELTA]t) is the discrete time series of x(t). The range of frequency is determined according to

(4) M[DELTA]f [less than or equal to] 1/2[DELTA]t.

One can plot the square of the absolute value of W(m[DELTA]f, b[DELTA]t) on the time-frequency plane as a scalogram on which this value is mapped to a gray scale value (see Figure 4).

[FIGURE 4 OMITTED]

The EEG data for 4.5 s for each task at the three levels of difficulty were analyzed using a wavelet transform, and the time-frequency characteristics at each task difficulty or work level were investigated. The number of EEG data at one wavelet transform was 4500 (4.5 s). The number of octave divisions and the final order for calculating wavelet coefficients were set to 3 and 12, respectively, so that the frequency bands were classified into [alpha] (8-12.4 Hz), [beta] (16-32 Hz), and [theta] (4-6.32 Hz) frequency bands. These frequency bands are determined by the choice of the number of octave divisions and the final order for calculating wavelet coefficients. In other words, these parameters were selected so that the frequency bands grossly correspond to the typical definitions of the theta, alpha, and beta frequency bands.

RESULTS

The relationship between task difficulty and reaction time is depicted in Figure 1. As the work level and session were within-subject factors, a fully within-subject design was used. A two-way (session by work level) repeated ANOVA conducted on reaction time revealed a main effect of only work level, F(2, 14) = 8.834, p < .01. A Fisher's protected least significant difference (PLSD) revealed significant differences for all of three combinations of two task levels. Mean reaction time was calculated on the basis of only correct trials. Figure 2 shows the relationship between task difficulty and the percentage correct. A similar two-way repeated ANOVA conducted on percentage correct revealed a main effect of only work level, F(2, 14) = 7.864, p < .01. A Fisher's PLSD revealed significant differences for the pairs (low, moderate), (low, high), and (moderate, high). Figure 3 compares the total NASA-TLX workload among the three levels of task difficulty. A Kruskal-Wallis nonparametric test conducted on the NASA-TLX total workload score revealed a significant main effect of work level (H = 7.261, p < .01). A one-way ANOVA conducted on the NASA-TLX total workload score revealed a significant main effect of work level, F(2, 14) = 18.392, p < .01. A Fisher's PLSD revealed significant differences for all three pairs of work levels.

[FIGURES 1-3 OMITTED]

In Figure 4, examples of scalograms (Fz) for each task difficulty are shown. The magnitude of power for each frequency is depicted on the time-frequency plane using a gray scale. The whiter scale represents a higher power. The three scalograms in Figure 4 show that the time when a maximum power (the area with the whitest scale in each figure) appears is delayed with increased task difficulty. As for Cz and Pz, similar tendencies were obtained. Based on the scalogram obtained from the wavelet analysis for EEG signals of 4.5-s duration at each task difficulty, the appearance times of the maximum power were obtained.

In Table 1, the relationship between task difficulty and the appearance time of the maximum power for three sites is shown. As a result of a three-way (session by work level by frequency band) ANOVA performed on appearance time, a significant main effect of only work level was detected, F(2, 14) = 14.286, p < .01. The three factors were all within-subject variables (a fully within-subject design was used). A Fisher's PLSD post hoc test revealed the following significant (p < 0.01) differences between the two conditions in each pair: (low, moderate), (low, high), and (moderate, high). As a result of a similar three-way (work level by frequency band) ANOVA performed on the mean appearance time for Cz and Pz, a significant main effect of only work level was detected: Cz, F(2, 14) = 23.624, p < .01; Pz, F(2, 14) = 14.318, p < .01. A Fisher's PLSD post hoc test detected the following significant (p < .01) differences between the two conditions in each pair for Cz and Pz: (low, moderate), (low, high), and (moderate, high).

The total power for the 4.5-s duration at each frequency band was also calculated at each work level using the results of the wavelet analysis. The total power at all frequency bands tended to increase as the work level increased. The relationship between task difficulty and total power is summarized in Table 2. As there are strong individual differences in the spectral power of EEG signals, the total power was normalized dividing the value by the maximum total power for each frequency band. A similar three-way (session by work level by frequency band) ANOVA revealed significant main effects of work level, F(2, 14) = 18.814, p < .01, and frequency band, F(2, 14) = 333.635, p < .01. A Fisher's PLSD post hoc test revealed the following significant (p < .01) differences between the two conditions in each pair: (low, moderate), (low, high), and (moderate, high) for work level, and ([alpha]-band, [theta]-band), ([alpha]-band, [beta]-band), and ([beta]-band, [theta]-band) for the frequency band. The total power also tended to increase with increased task difficulty.

As a result of a similar ANOVA carried out on the total power for Cz and Pz, significant main effects of work level--Cz, F(2, 14) = 5.058, p < .05; Pz, F(2, 14) = 4.835, p < .05--and frequency band--Cz, F(2, 14) = 402.635, p < .05; Pz, F(2, 63) = 198.681, p < .05--were detected. A similar post hoc test performed on the total power for Cz and Pz detected significant (p < .01) differences of all combinations of work levels and frequency bands.

The results of FFT analysis for the same data as the wavelet analysis are shown in Table 3. The analysis was conducted according to Gevins et al. (1998). It must be noted that only the [theta]-, [alpha]-, and [beta]-frequency bands were equal to those in the wavelet analysis. As with the total power in Table 2, the power was normalized by dividing the value by the maximum power for each of three frequency bands. Unlike the results of wavelet analysis, the normalized power cannot differentiate between the low and middle work levels, although the normalized power differed between the higher work level and the low or moderate work level. A similar three-way (session by work level by frequency band) ANOVA carried out on the power revealed significant main effects of work level, F(2, 14) = 3.904, p < .01, and frequency band, F(2, 14) = 196.247, p < .01. A Fisher's PLSD post hoc test revealed the following significant (p < .01) differences between the work level conditions: (low, high) and (moderate, high).

DISCUSSION

Reaction time increased as task difficulty increased (Figure 1). In accordance with this, the percentage correct decreased with increased task difficulty (Figure 2). The rating score of each item in the NASA-TLX also tended to increase as task difficulty increased (Figure 3). The experimental tasks induced different cognitive workloads in the participants. The performance data and results of the NASA-TLX validate that the experimental task was proper and that the participant sensed that the workload was heavier when the task was more difficult.

In Figure 4, the white scale corresponds to the area with a higher spectral power. The power spectral component changes with time. The time until the maximum power appears on the time-frequency scale is delayed as the task difficulty is increased. The appearance time seems to correspond well with the change of task difficulty. The appearance times of the three frequency bands ([theta], [alpha], and [beta]) were calculated for each task difficulty using the scalogram of each participant. As shown in Table 1, the appearance time increased as task difficulty increased. With an increase in task difficulty, reaction time was prolonged and heavier workload was induced in the participant. This phenomenon was reflected in the appearance time obtained from the wavelet analysis of EEG signals, a measure that seems to be more objective than the performance and psychological data. This is indicative of the effectiveness of the appearance time as a means to evaluate mental workload.

The relationship between the total power and task difficulty during a 4.5-s analysis interval for each frequency band is shown in Table 2. The total power, as well as the appearance time, corresponded well with the change of task difficulty. As task difficulty increased, the total power for all three frequency bands tended to increase. This shows that the central nervous system is activated when a heavier cognitive workload is imposed on a participant. On the basis of the appearance time and total power extracted from the wavelet analysis of EEG signals, one can discriminate among cognitive task difficulties. The two parameters are as reliable as that obtained from the traditional spectral analysis of biological signals (Gevins et al., 1997, 1998; Gevins, Zeitlin, Doyle, Schafer, et al., 1979; Gevins, Zeitlin, Doyle, Yingling, et al., 1979) and can be applicable in a wider range of work environments in human-computer interactions. It must be noted that the two measures showed different patterns of change as a function of both task difficulty and frequency band. The appearance time showed a pattern of increasing as a function of task difficulty. However, the appearance time did not differ among the three frequency bands. The total power increased with the increase of task difficulty and differed among the three frequency bands.

Gevins et al. (1998) found [beta] power to be unaffected by task difficulty. The [theta] activity increased with increasing task difficulty and corresponded with the result in this study. The [alpha] activity decreased from the lowest to the highest task difficulty. The power spectrum, assuming that EEG signals are stationary, shows the global frequency characteristics during an analysis interval. The power obtained by FFT analysis for all three frequency bands was affected by the work level (Table 3). The power tended to increase with increasing task difficulty, although there was no statistical difference in power between the low and moderate work levels. The results of only the [theta] band corresponded with Gevins et al. (1998). Unlike the results of wavelet analysis, the power obtained by FFT could not definitely differentiate among the three work levels.

The wavelet analysis tracks the change of time-frequency characteristics. As the total power in this study is not necessarily the same as the [alpha], [beta], and [theta] power in Gevins et al. (1998), a direct comparison might not be proper. In general, [alpha] and [beta] rhythms are characteristically manifested in the form of increased alertness caused by higher mental activity (Bechtereva, 1981; Gardner; 1975). In view of this, the increasing [alpha]- and [beta]-band total power (Table 2) or power (Table 3) as a function of task difficulty might be reasonable. As the change of [alpha]- and [beta]-band total power or power as a function of task difficulty showed a tendency different from that in Gevins et al. (1998), future research should verify the validity of total power. The appearance time for each task difficulty was nearly constant irrespective of frequency band, which is consistent with past findings. The reason for this should also be addressed in future research. Therefore, using the appearance time as an evaluation measure would be preferable at the present stage.

The results provide evidence that as task difficulty increases, there is increased neocortical activity and a higher rating of mental workload. increased activity in the central nervous system is further supported by an increased appearance time extracted from the wavelet analysis of EEG signals. Although it is not easy to definitely define mental workload and measure it reliably, this must be approached with multidimensional indices. It is common to consider that the mental resources for perceptual and cognitive processes are different from and independent of those for motor processes. Keeping this in mind, it must also be noted that the activity of the motor system selectively modulates EEG features such as somatomotor mu rhythm (Pfurtscheller, Neuper, & Berger, 1994). Therefore, caution is needed when evaluating mental resources allocated to mental task performance using EEG signals, so that the effects of perceptual and cognitive systems can be separated from those of motor systems. The selection of an experimental task or the structure of the mental resources allocated to task performance may be a crucial factor in the empirical evaluation of mental workload.

The finding that the appearance time and the total power extracted from wavelet analysis are useful for evaluating mental workload would be applicable and generalizable in work environments that require relatively greater perceptual and cognitive resources, such as arithmetic and memory tasks. Compared with HRV for the evaluation of mental workload (Akselrod et al., 1981; Baselli & Cerutti, 1985; Baselli et al., 1987; Cerutti et al., 1988; Murata, 1994), the respiration interval might be less influential when EEG signals are used to evaluate mental workload. In this study, the respiration interval was roughly controlled, by way of precaution, using a respiration interval between 3 and 4 s. Future research should explore the limitation of the applicability and generalizable nature of the proposed method in more detail. The effects of the respiration interval on EEG signals might also need to be further explored.

Using time-frequency characteristics obtained by wavelet analyses, one can classify differences in cognitive task difficulty in a shorter period than is possible with traditional approaches. Classification of cognitive task difficulty using ERP measurements (Johnson & Donchin, 1980; Kramer et al., 1985; Ullsperger et al., 1988) requires a great deal of time to average about 50 EEG waveforms elicited by a rare stimulus, which appears with a probability of 0.2. This approach also induces a monotonous feeling in the participant and might consequently degrade the accuracy of classification. Gevins et al. (1998) tried to assess working memory load during computer use with neural network pattern recognition applied to EEG spectral features. They succeeded in the classification of low and high load levels with an accuracy of more than 95%. However, it is still difficult to classify three load levels (low, moderate, and high) clearly and definitely.

There seem to be some limitations in the classification of load levels based solely on the power spectrum of EEG signals. The method proposed in this study, however, can differentiate cognitive task difficulty definitely, and in a shorter period, based on the appearance time and total power obtained. This difference can be attributed to the higher sensitivity of time-frequency characteristics to changes in cognitive task difficulty. The differences in cognitive task difficulty appear clearly in the time-frequency characteristics in Figures 4a through 4c. The function of the central nervous system differs according to cognitive task difficulty. In other words, increasing cognitive task difficulty seems to delay the time at which the central nervous system works most actively. This can be detected not by using traditional approaches but by using an approach that makes use of the time-frequency characteristics of EEG signals. In view of this, the appearance time would be the more preferable of the two measures proposed in this study.

It is important to develop multidimensional indices that can differentiate between loadings from the perceptual and cognitive system and those from the motor system. Wavelet analysis would be helpful in tracking, at every moment, the change of frequency characteristics of EEG signals, which has not been realized using traditional spectral analysis. As one can extract the appearance time from the wavelet analysis, the development of multidimensional indices for the evaluation of mental workload needs the consideration of a time factor. Overcoming the aforementioned problems, the method could be promising for monitoring and evaluating mental workload in human-computer interaction that requires relatively greater perceptual and cognitive loadings or resources.

In conclusion, the results indicate that a wavelet transform of EEG signals is a promising measure for the evaluation of mental workload. Using wavelet analysis and extracting parameters such as the total power and the appearance time, one can reliably evaluate mental workload. Future research should explore and verify the validity of proposed parameters under more realistic human-computer interaction environments.

REFERENCES

Aasman, J., & Mulder. L. J. M. (1987). Operator effort and measurement of heart-rate variability, Human Factors. 29. 161-170.

Akselrod, S., Gordon. D., Ubel, F. A., Shannon, D. C., Barger, A. C., & Cohen, R. J. (1981). Power spectrum analysis of heart rate fluctuation: A quantitative probe of beat-to-beat cardiovascular control. Science. 215, 220-222.

Bartoli, S., Baselli, G., & Cerutti, S. (1985). AR identification and spectral estimate applied to the R-R interval measurement. International Journal of Biomedical Computing. 16, 201-215.

Baselli. G., & Cerutti. S. (1985). Identification technique applied to processing of signals from cardiovascular systems. Medical Informatics, 10, 223-235.

Baselli, G., Cerutti. S., Civardi, S., Lombardi, F., Malliani, A., Merri, M., et al. (1987). Heart rate variability signal processing: A quantitative approach as an aid to diagnosis in cardiovascular pathologies. International Journal of Bio-Medical Computing, 20, 51-70.

Bechtereva, N. R (1981). Neurophysiological correlates of mental process in man. In N. P. Bechtereva (Ed.), Psychophysiology today and tomorrow, (pp. 11-21). New York: Pergamon Press.

Cerutti, C., Fortis, G., Libeart, D., Baselli. G., & Civardi. S, (1988). Power spectrum analysis of heart rate during a mental arithmetic task. Journal of Ambulatory Monitoring, 1, 241-250.

Coifman, R. R., & Wickerhauser, M.V. (1993). Wavelets and adapted waveform analysis. Wellesely, MA: AK Press.

Daubechies, I. (1988). Orthonormal bases of compact supported wavelets. Communications of Pure and Applied Mathematics. 41, 909-996.

Daubechies, I. (1990). The wavelet transform, time-frequency localization and signal analysis. IEEE Transactions on Information Theory, 56, 961-1005.

Gardner, E. (1975). Fundamentals of neurology. Philadelphia: Saunders.

Gevins, A., Smith. M. E., Leong, H., McEvoy, L., Whitfield, S., Du, R., et al. (1998). Monitoring working memory load during computer-based tasks with EEG pattern recognition methods. Human Factors, 40, 79-91.

Gevins. A., Smith, M. E., McEvoy. L., & Yu. D. (1997). High resolution EEG mapping of cortical activation related to working memory: Effects of task difficulty, type of processing, and practice. Cerebral Cortex, 7, 374-385.

Gevins, A., Zeitlin, G. M., Doyle, J. C., Schafer, R. E., & Callaway, E. (1979). EEG patterns during "cognitive" tasks: II. Analysis of controlled tasks. Electroencephalography and Clinical Neurophysiology, 47, 704-710.

Gevins, A., Zeitlin, G. M., Doyle, J. C., Yingling, C. D., Schafer, R. E., Callaway, E., et al. (1979). EEG correlates of higher cortical functions. Science, 203, 665-668.

Gundel, A., & Wilson, G. F. (1992). Topographical changes in the ongoing EEG related to the difficulty of mental tasks. Brain Topography, 5, 17-25.

Hart. S. G., & Staveland, L. E. (1988). Development of NASA-TLX (Task Load Index): Results of empirical and theoretical research. In P. A. Hancock & N. Meshkati (Eds.), Human mental workload (pp. 159-185). Amsterdam: North-Holland.

Hill, S. G., Iavecchia, H. P., Byers, J. C., Bittner, A. C., Zaklad, A. L., & Christ, R. E. (1992). Comparison of four subjective workload rating scales, Human Factors, 34, 429-439.

Humphrey, D., & Kramer, A. F. (19941. Toward a psychophysiological assessment of dynamic changes in mental workload. Human Factors, 36, 5-26.

Johnson, R., & Donchin, E. (1980). P300 and stimulus categorization: Two plus one is not so different from one plus one. Psychophysiology, 17, 167-178.

Kramer, A. F., Wickens, C. D., & Donchin. E. (1985). Processing of stimulus properties: Evidence for dual-task integrality. Journal of Experimental Psychology: Human Perception and Performance. 11, 393-408.

Luczak, H., & Laurig, W. (1973). An analysis of heart rate variability; Ergonomics, 16, 107-116.

Meyer, Y., & Ryan, R. D. (1993). Wavelets--Algorithms and applications. Philadelphia, PA: Society for Industrial and Applied Mathematics.

Murata, A. (1994). Experimental discussion on measurement of mental workload--Evaluation of mental workload by HRV measures. Institute of Electronics, Information, and Communication Engineers Transactions of Fundamentals of Electronics, Communications, and Computer Sciences, E77-A, 409-416.

Okogbaa, O. G., Shell, R. L., & Filipusic, D. (1994). On the investigation of the neurophysiological correlates of knowledge worker mental fatigue using the EEG signal. Applied Ergonomics, 25, 355-365.

Pfurtscheller, G., Neupeur, C., & Berger, J. (1994). Source localization using event-related desynchronization (ERD) within the alpha band, Brain Topography, 6, 269-275.

Tsuji, H., & Mori, H. (1994). New analysis of HRV through wavelet transform. International Journal of Human-Computer Interaction. 6, 205-217.

Ullsperger, P., Metz, A.-M., & Gille, H.-G. (1988). The P300 component of the event-related brain potential and mental effort. Ergonomics, 31, 1227-1157.

Wilson, G. F., & Fisher, F. (1995). Cognitive task classification based upon topographic EEG data. Biological Psychology, 40, 239-250.

Young, R. K. (1993). Wavelet theory and its applications. Boston, MA: Kluwer Academic.

Atsuo Murata is a professor in the Department of Computer and Media Technologies at Hiroshima City University. He received his Ph.D. in engineering in 1987 from Osaka Prefecture University.

Date received: September 5, 2003

Date accepted: May 13, 2004

Address correspondence to Atsuo Murata, Hiroshima City University, Department of Computer and Media Technologies, 3-4-1, Ozukahigashi, Asaminami-ku, Hiroshima 751-3194, Japan: murata@cs.hiroshima-cu.ac.jp.
TABLE 1: Comparison of Mean Appearance Time (in Seconds) Among Levels of
Task Complexity, Frequency Bands, and Sites (Wavelet Analysis)

               [theta] Band                      [alpha] Band

        Low      Moderate     High        Low      Moderate     High

Fz     0.726      0.949       1.204      0.761      0.959       1.241
      (0.161)    (0.248)     (0.447)    (0.177)    (0.231)     (0.467)
Cz     0.952      1.203       1.485      0.936      1.254       1.464
      (0.171)    (0.195)     (0.342)    (0.156)    (0.198)     (0.384)
Pz     1.102      1.274       1.564      1.132      1.355       1.532
      (0.135)    (0.251)     (0.238)    (0.176)    (0.234)     (0.312)

                [beta] Band

        Low      Moderate     High

Fz     0.779      0.979       1.252
      (0.205)    (0.239)     (0.472)
Cz     0.925      1.237       1.407
      (0.102)    (0.245)     (0.239)
Pz     1.096      1.238       1.463
      (0.139)    (0.265)     (0.305)

Note. Standard deviations are in parentheses.

TABLE 2: Comparison of Mean Total Power Among Levels of Task
Complexity, Frequency Bands, and Sites  (Wavelet Analysis)

               [theta] Band                   [alpha] Band

        Low      Moderate    High      Low      Moderate     High

Fz     0.688      0.793        1      0.554      0.609       0.744
      (0.245)    (0.258)             (0.148)    (0.196)     (0.206)
Cz     0.743      0.876        1      0.605      0.638       0.798
      (0.141)    (0.156)             (0.137)    (0.104)     (0.273)
Pz     0.715      0.836        1      0.598      0.625       0.789
      (0.175)    (0.164)             (0.201)    (0.146)     (0.176)

                [beta] Band

        Low      Moderate     High

Fz     0.339      0.372       0.452
      (0.104)    (0.134)     (0.127)
Cz     0.349      0.383       0.487
      (0.136)    (0.154)     (0.143)
Pz     0.315      0.398       0.485
      (0.175)    (0.165)     (0.135)

Note. Standard deviations are in parentheses.

TABLE 3: Comparison of Mean Power Among Levels of Task Complexity,
Frequency Bands, and Sites (FFT)

               [theta] Band                   [alpha] Band

        Low      Moderate    High      Low      Moderate     High

Fz     0.846      0.854        1      0.624      0.632       0.775
      (0.215)    (0.148)             (0.135)    (0.181)     (0.167)
Cz     0.798      0.807        1      0.647      0.651       0.735
      (0.151)    (0.205)             (0.110)    (0.162)     (0.184)
Pz     0.813      0.824        1      0.598      0.575       0.713
      (0.190)    (0.251)             (0.108)    (0.130)     (0.213)

                [beta] Band

        Low      Moderate     High

Fz     0.405       0.41       0.468
      (0.105)    (0.139)     (0.192)
Cz     0.423      0.416       0.475
      (0.142)    (0.105)     (0.201)
Pz     0.391      0.403       0.438
      (0.109)    (0.168)     (0.257)

Note. Standard deviations are in parentheses.
COPYRIGHT 2005 Human Factors and Ergonomics Society
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2005 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Murata, Atsuo
Publication:Human Factors
Geographic Code:1USA
Date:Sep 22, 2005
Words:6030
Previous Article:Target acquisition with UAVS: vigilance displays and advanced cuing interfaces.
Next Article:Heuristic automation for decluttering tactical displays.
Topics:

Terms of use | Privacy policy | Copyright © 2020 Farlex, Inc. | Feedback | For webmasters