Receive Package and Display

For display purpose, we expected the program received transmission digital data package continuously via Bluetooth. The data package format follows MSP430 encoding which explained in Table 5 of session 2.3.3.3 B. which is including one byte of header, one byte of data information, and 16 channel data. The program recognizes out package header through checking receiving data from byte to byte. On the other words, this program owns itself communication protocol, which once finding out common header, begins to decode certain package information and the following fix length of data. Table shows package protocol and function (2-1) shows decode thesis.

_ _

_ 2^{data value}, _ 2^{data value}

Sampling rate= Channel number= (2-1) Table 7: Header of package

Header Data information

f f Sampling rate Channel number Table 8: data information of package

data value 0 1 2 3 4 5 6 7 8 9 10

B. Display Digital Data to Analog Signal

Actual EEG signal= ×Power −Virtual GND potential

12

( _ 3) 0.5

2 data value

= × −

The program already defaults ADC resolution is 12 bits and isn’t adapted by user interface. As for the actual potential value is not influence the method of display.

After the program receive the 0-4096 range of data value, it just plots multi-channels data values in width-required frame, and shows the curve length according to sampling rate from package information and the number of display second from user interface.

2.3.5.2 Save Data as a File

Display program which was installed in the computer continuously receives digital data packages one by one via Bluetooth if it is started. In the beginning, user interface would ask user whether to save data as a file or not. If the save request is set, the program not only displays real-time converted analog signal but also saves received digital data as a TXT-format file simultaneously. In the TXT file, data package is saved as a row value once, and each channel data is saved as column value.

The first column means the first channel data, and the second column means the second channel ones, and so on. From left to right put each channel data in order and the final column means the 16-th channel data or event value if user needs.

Incidentally referring, the program record received event data with the rate of receiving. Program would get event file name and event input comport from user interface. So that every time the program receives one package, it synchronously checks certain input comport whether receiving event information or not. If event

value appears, the program recognizes it and save a value in final column behind 16 columns. However, if event value doesn’t appears, the program just save zero value in this event column. As following, Fig. 2-29 shows data appearance in TXT file.

a Package received once

Event column if needed

certain Channel data received by time

Left to right: Channel 1 to 16

Fig. 2- 29: TXT file contexts 2.3.5.3 User’s Guide

Fig. 2- 30: User interface with opening window

As Fig. 2-30 shown, the 1^st row mentions certain Bluetooth module using now.

After opening this program, user can click “Refresh” button to search existing Bluetooth modules in finite distance. The 2^nd row asks user that this program is going to display how much time for EEG real-time signal. The 3^rd row asks user whether to save receiving data as a TXT file or not. If yes, please select a new file name. The 4^th row asks user whether simultaneously receiving event value via a specified comport or not. The 5^th row asks user whether to let program automatic checking the synchronization between event and channel data. After finished above request menu, user can click “Beginning” button to start this program. In addition to, to click “Quit”

button stops this ongoing program. Moreover, Fig. 2-31 shows the window once starting to receive EEG data.

Fig. 2- 31: Appearance of display Program within receiving process

2.4 Experiment: P300 and Oddball task

For system verification purpose, we need a standard cognitive experiment to make sure our proposed mobile and wireless system can measure tiny EEG activity even though Event-Related Potential (ERP). Thus, we introduce P300 and Oddball task in this session progressively.

2.4.1 P300

Event-related potential (ERP) refer to averaged EEG responses that are time-locked to more complex processing of stimuli; this technique is used in cognitive science, cognitive psychology, and psychophysiological research. An event-related potential (ERP) is any measured brain response that is directly the result of a thought or perception. More formally, it is any stereotyped electrophysio- logical response to an internal or external stimulus.[24]

Fig. 2- 32: (a) P300 [24] (b) Context updating theory of P300 [18]

The P300 (Fig. 2-32 (a)) is a positive component of the event-related potential (ERP) that peaks 300ms or more (up to 900 ms) after a stimulus. Unlike some of the earlier evoked potentials, it is supposed to be an “endogenous” component in the

sense that it depends very much on the processing of the stimulus context and levels of attention and arousal (Polich and Kok 1995). The P300 has commonly been investigated with “oddball” paradigms, in which occasional relevant (“target”) stimuli have to be detected in a train of frequent irrelevant “non-target” or “standard” stimuli.

Such oddball paradigms reliably yield P300 responses with a parietocentral scalp distribution to target compared to standard stimuli irrespective of stimulus (visual, auditory, somatosensory) or response (button press, counting) modality. Interestingly, P300 responses are also observed when trains of regular stimuli are interrupted by stimulus omissions, which underlines the endogenous nature of this component.

The amplitude of the P300 increases with lower probability and higher discrimination of targets. Its latency increases when targets are harder to discriminate from standards but not when response times increase for other reasons. P300 latency is thus an attractive tool to separate the mental chronometry of stimulus evaluation from response selection and execution (Coles and others 1995). The P300 is widely believed to be a neural signature of the mechanisms required to change the mental model of the environment to make an appropriate response (Polich 2003). In terms of classical cognitive domains, both attention (selecting the deviant stimulus from the train of irrelevant stimuli) and working memory (supporting this process by maintaining the features of the standard stimulus for comparison) seem to be involved.

However, the higher amplitude of the P300 for easily discriminated targets (that is, when demand on working memory should be low) and the lower amplitude in tasks with high memory load (Kok 2001) indicate that the interplay between attention and working memory in the generation of the P300 is not straightforward and certainly not simply additive.[18]

Fig. 2-32 (b) shows schematic illustration of the P300 context-updating model (Polich, 2003). Stimuli enter the processing system and a memory comparison process is engaged that ascertains whether the current stimulus is either the same as the previous stimulus or not (e.g. in the oddball task, whether a standard or a target stimulus was presented). If the incoming stimulus is the same, the neural model of the stimulus environment is unchanged, and sensory evoked potentials (N100, P200, and N200) are obtained after signal averaging. If the incoming stimulus is not the same and the subject allocates attention resources to the target, the neural representation of the stimulus environment is changed or updated, such that a P300 (P3b) potential is generated in addition to the sensory evoked potentials. [18]

2.4.2 Oddball Task

P300 occurs when a stimulus is presented. Oddball task is just one of method to detect it. Besides, past researchers also designed single-stimulus and three-stimulus experiment to study the characteristic of P300. As following Fig. 2-33 shown, single-stimulus experiment induce single P300, oddball experiment induce a larger amplitude of P300 when stimulus discrimination occurred, and three-stimulus experiment induce P3a and P3b when distracter stimulus and difficult stimulus discrimination occurred respectively.

As Fig. 2-33, schematic illustration of the single-stimulus (top), oddball (middle), and three-stimulus (bottom) paradigms, with the elicited ERP from the stimuli of each task at the right (Polich and Criado, 2006). The single-stimulus task presents an infrequent target (T) in the absence of any other stimuli. The oddball task presents two different stimuli in a random sequence, with one occurring less frequently than the other does (target=T, standard=S). The three-stimulus task is similar to the oddball with a compelling distracter (D) stimulus that occurs infrequently. In each task, the

subject is instructed to respond only to the target and otherwise to refrain from responding. The distracter elicits a P3a, and target elicits a P3b (P300). [20]

Fig. 2- 33: Schematic illustration of the single-stimulus (top), oddball (middle), and three-stimulus (bottom) paradigms [20]

Fig. 2- 34: Experimental scene description [21]

In past study, Marco D. Comerchero and John Polich designed an oddball task and investigated P300 of different position. Fig. 2-34 summarizes the stimulus properties for the visual modalities. EEG activity was recorded during 2 blocks, each of which consisted of 350 stimulus presentations that lasted approximately 12 min.

Stimuli were defined as target, non-target, and standard and presented with probabilities of 0.10, 0.10, and 0.80, respectively. Task conditions were defined according to the level of perceptual difficulty of the target/standard discrimination, such that for each modality subjects were presented with one Easy and one difficult condition. The task in all conditions was to respond to the target stimulus by pushing a mouse button with the right index finger as quickly and accurately as possible.

Response time and error rates were recorded. All subjects were given a practice block consisting of 15 stimulus trials before each condition. Modality order was counterbalanced across subjects, but the Easy task was always presented first for each modality to promote successful task performance in the subsequent difficult condition.[21]

Fig. 2- 35: Grand averaged ERP from visual modalities for each task difficulty, stimulus type, and recording site (n = 16) [21]

Fig. 2-35 presents the grand average ERP from the target, standard, and non-target auditory and visual stimuli in the Easy and difficult task conditions. For the Easy tasks, target stimuli elicited P300 components that were largest at the parietal electrode. Non-target stimuli elicited P300 components that were similar in morphology to those elicited by target stimuli but with appreciably smaller amplitude across all electrode sites. For the difficult tasks, target stimuli elicited P300 components that exhibited smaller amplitudes and longer latencies than those from the Easy tasks. Non-target stimuli elicited P300 components that were larger and earlier at the frontal and central electrodes than those from the target stimuli. Target P300 amplitude was larger than the non-target amplitude at the parietal electrode.[21]

P300 scalp distribution is defined as the amplitude change over the midline electrodes (Fz, Cz, and Pz), which typically increases in magnitude from the frontal to parietal electrode sites (Johnson, 1993).[21] As Fig shown, we can get a conclusion of the amplitude of Pz is the largest, the medium amplitude is in Cz, and the smallest amplitude is in Fz.

In addition, what about the relationship between target-to-target interval and P300 amplitude? The following fig. presents the results. We can get information that no matter using TT, NT, NNT, or NNNT experiment, the major reason of deciding the P300 amplitude is target-to-target interval. And the largest amplitude detected in about 12-second target-to-target interval.

Fig. 2- 36: Relationship between P300 and target-to-target interval (TTI) [21]

As Fig. 2-36, P300 amplitude plotted as a function of target-to-target interval (TTI) for the target (T) stimulus in an oddball task across sequences of preceding non-target (N) standard stimuli. The legend defines the symbols used to depict various non-target and target sequences. The subject is instructed to respond only to the target stimulus. P300 amplitude increases independently of local sequence and global target probability. The regression lines reflect curvilinear best fit for a second order polynomial. Similar results have been found for the single-stimulus paradigm when only target stimuli are presented (Gonsalvez et al., 2007). Adapted from Gonsalvez and Polich (2002) with permission of the authors and Blackwell Publishing (Copyright 2002).[21]

To summarize the above researches, we get four important points: (1) oddball task includes standard stimulus and target stimulus, both inducing different amplitude of ERP, (2) P300 amplitude is larger in easy stimulus discrimination than difficult one, (3) the amplitude change over the midline electrodes (Fz, Cz, and Pz), which typically increases in magnitude from the frontal to parietal electrode sites, and (4) target-to-target interval also effect P300 amplitude. By referencing to these, we design an oddball experiment in this study and then also verify P300 pattern could be measured by our proposed system.

Chapter 3 Experiment Results and Discussion

In this chapter, we verify the reliability of our proposed EEG system step by step.

In order to the whole multi-channels mobile and wireless EEG system is consisting of many independent modules which need to be verified respectively. The method of the whole system verification (Fig. 3-1) is mainly separated four sessions: (1) system verification of simulated signal, (2) circuitry and program test compared to reference system, (3) sensor basic test, and (4) performance test of sensors in Oddball task.

First, 16-channel EEG device is inputted simulated signals. Check the correlations in time and frequency domains from TXT file saved by back-end record program. The second, our proposed EEG acquisition circuitry and program is compared by reference system for comparison in time and frequency domains through three-action experiment. Next, dry and wet sensors would be respectively placed the surrounding positions on subjects’ head to verify whether dry sensor is suitable for long-term monitoring and good convenience in our EEG system. Moreover, the four-action experiment verifies four kinds of characteristic performances in our EEG system. In this part, sensor would be verified the less difference between three subjects. Finally, two kinds of sensors are also really applied in Oddball task, and we verify the dry sensor indeed measures tiny EEG activity like Event-Related Potential (ERP) from ten subjects.

Fig. 3- 1: System verification overview

3.1 System Verification of Simulated Signals

The method for system verification of simulated signals is that our EEG acquisition circuitry catches standard simulated signals produced by function generator, and then the signals are amplified, filtered, and transmitted to back-end EEG display program via Bluetooth. We compare record data saved by program and standard simulated signals created by MATLAB. Analysis technique is computing data correlation in time and frequency domains respectively. By this way, we effectively get how good performance our EEG system can be.

3.1.1 Performance Test in Time Domain

As Fig. 3-2 shown, function generator produced 5Hz simulated signal to circuitry, and program received amplified digital data like blue curve. In addition, red curve created by MATLAB was standard 5Hz simulated signal. Fig. shows 20-second data and also masks computed time-domain correlation in each second. The results illustrate every one-second correlation is high above 99.5%.

Fig. 3- 2: Result of 5Hz simulated signal test

Fig. 3- 3: result of non-linear simulated signal test

Apart from the above-mentioned, we also tried to establish a non-linear simulated signal through function generator, and equally the signal was record by our program. To make a comparison between the pre-designed data and received data, we found out our proposed EEG system really record the features of certain non-linear signal in each time. Like Fig. 3-3, blue curve is pre-designed non-linear simulated signal, and red curve is record data by our system. Because of our system was created low-sampling rate of 125Hz, record data automatically filtered the noise over 125Hz.

Hence, red curve is more clear than blue curve. Similarly, the result showed the best performance of simulated signal test in time domain.

3.1.2 Performance Test in Frequency Domain

We used function generator to produce 5Hz, 10Hz, 15Hz, and 20Hz simulated frequency respectively. Then FFT comparison verified the performance in frequency domain. As Fig. 3-4 shown, each specific frequency is correctly received by system.

Fig. 3- 4: Results of different frequency test

3.2 Circuitry and Program test compared to reference

Fig. 3- 5: Diagram of circuitry and program test compared to reference system

The portable and wireless EEG acquisition circuitry and display program combine EEG preprocessing and recording data functions respectively. In advance, these could be compared with international standard EEG measurement instrument like Neuroscan system. Neuroscan system more or less includes electrode cap which need conductive gels, SynAmps2 amplifier, scan software. In this verification, we use the same sensor to make sure that the comparison can’t be affect due to the difference of dry sensor and wet electrode. To choose dry sensor rather than wet electrode is in order to guarantee the performance results are not owing to wet electrode. Moreover, SynAmps2 owns itself specification of circuitry, and we can’t get this detail information from Neuroscan Company. Besides, our proposed system receives data

via wireless path, but Neuroscan system receives data through wire path.

Consequently the goal of verification just emerges the comparison results rather than how good performance among two systems.

Fig. 3-5 shows diagram of system comparison. Neuroscan SynAmps2 uses 200-Hz sampling rate to measure EEG activity, and our 16-channel EEG device use 125-Hz ones. The scan software interface provides menu to select suit sampling rate, so we select the most similar one- 200Hz. the file format of record data is different for two systems, but this problem could be improve in analysis stage. Before going to comparison, record data by Neuroscan system is first re-sampled to 125 Hz.

3.2.1 Participant

It is sufficient that only one subject participles the comparison test. This subject is a female. She is 24 years old and trends to have a large number of hairs.

3.2.2 Experiment Procedure and Presentation

Human brain emerges EEG activity mixing different frequencies in different pattern cognition state. Thus, we can take advantage of EEG unique characteristic to verify the system performance in different frequency band. In this experiment, we assigned the participant to behave three kinds of actions which are “Blink”, “Tooth”, and “Normal”. “Blink” action, which is referred to EOG signal, verifies the lower frequency band about 0.2 - 5 Hz. “Tooth” action, which is referred to EMG signal, verifies the higher frequency band about 10 – 25 Hz. “Normal” action verifies frequency band about 0.2 – 25 Hz. In this procedure of experiment, we required the participant to blink once per second within “Blink” command occurred, to grind the molar with uninterrupted within “Tooth” command occurred, and to do general action naturally within “Normal” command occurred. Among Normal state, EOG like eye-movement and EMG like muscle-movement at chin position are randomly

appeared to cover clear EEG signal. Thus, EOG and EMG are ordinary considered no-use activity and even interfering more important EEG feature in fact. However, system comparison in this experiment can use the characteristic of EOG and EMG to check lower and higher frequency bands respectively. Fig. 3-6 shows the space of experiment.

Fig. 3- 6: Experiment environment and setting view

Beginning of experiment, the participant sat on a chair motionlessly and was asked to keep facing on the front screen and followed occurred commands by presentation. The experiment includes four sections, each of about 10-miniute duration. Each section includes thirteen trials, each of forty-five-second duration. The participant can take a rest among sections. Fig. 3-7 shows procedure of presentation in one trial.

Fig. 3- 7: Experiment procedure in one trial

Referencing to 10-20 system (Guideline for Standard Electrode Position Nomenclature, 2006), we choose the edge and center positions (Fig. 3-8) to place dry sensor. The forefront side is FP1. The most left side is T3. The top side is CZ. The most right side is T4. The most behind side is OZ. In addition, the participant is also placed an electrode behind ear as reference potential and an electrode on G as ground

Referencing to 10-20 system (Guideline for Standard Electrode Position Nomenclature, 2006), we choose the edge and center positions (Fig. 3-8) to place dry sensor. The forefront side is FP1. The most left side is T3. The top side is CZ. The most right side is T4. The most behind side is OZ. In addition, the participant is also placed an electrode behind ear as reference potential and an electrode on G as ground