This paper presents an investigation aimed at drastically reducing the processing burden required by motor imagery brain-computer interface (BCI) systems based on electroencephalography (EEG). In this research, the focus has moved from the channel to the feature paradigm, and a 96% reduction of the number of features required in the process has been achieved maintaining and even improving the classification success rate. This way, it is possible to build cheaper, quicker, and more portable BCI systems. The data set used was provided within the framework of BCI Competition III, which allows it to compare the presented results with the classification accuracy achieved in the contest. Furthermore, a new three-step methodology has been developed which includes a feature discriminant character calculation stage; a score, order, and selection phase; and a final feature selection step. For the first stage, both statistics method and fuzzy criteria are used. The fuzzy criteria are based on the S-dFasArt classification algorithm which has shown excellent performance in previous papers undertaking the BCI multiclass motor imagery problem. The score, order, and selection stage is used to sort the features according to their discriminant nature. Finally, both order selection and Group Method Data Handling (GMDH) approaches are used to choose the most discriminant ones.
Brain-computer interface (BCI) systems capture brain signals and decode them with the purpose of interacting with external devices without any muscular or physical intervention. Well-known examples are motor imagery tasks due to their importance in applications for severely motor impaired people. Likewise, other patterns can also be recognized within the brain signals, including word generation or object rotation. These patterns can be transformed to distinguishable signals and then to external commands or actions [
Technologically, most of the BCI mechanisms are based on electroencephalogram (EEG) techniques, where the sensors detecting the electric potentials originated by the neurons are placed on the scalp of the user [
According to how the brain signals get activated, two different paradigms can be distinguished [
Because the recorded brain signals are so small in amplitude, EEG devices in particular present a very low signal to noise ratio (SNR). For this reason, any interference coming from sources such as eye movement, eye-blink, muscular movements, teeth clash, or the heart rhythm deeply affects the quality of the measured signal, which can prevent the decoding system from properly recognizing the intention of the user. As a consequence, an effort to improve the spatial filtering methods [
In recent years, there has been increasing interest in minimizing the number of channels and features used by the classification algorithms. Yang et al. [
There are a large number of published studies describing different approaches to feature and channel selection. These approaches comprise both wrapper and filter methods of feature selection. The most popular methods are Genetic Algorithms (GA) [
In [
Although extensive research has been carried out on feature selection, most of the available research has focused on reducing the number of channels required instead of the number of individual features. Also, no single study exists which adequately covers the result of implementing Statistical and Fuzzy approaches.
In Cano-Izquierdo et al. [
Later work [
The data processing on the BCI Competition data sets is always off-line. If the methods included on the literature were to be applied on live applications, the time constraints to produce a prediction would be a major issue to address. For instance, for the Data Set V, it is necessary to calculate the PSD function for 8 sensors and 12 frequency bands (96 features) and then apply the recognition logic 16 times per second. Moreover, there is a requirement of producing a prediction every 0.5 seconds. This computational burden requirement is not easily accommodated even on today’s PCs. For on-line applications, reduction of the number of features to process is necessary.
This paper introduces a new methodology to choose the most relevant features using different approaches, being the statistic properties of the data or the relationship between the fuzzy categories which are generated on a S-dFasArt model. These methods have been applied to the Data Set V available for BCI Competition III [
The remainder of this paper is organized as follows. Section
The work presented in this paper is based on the Data Set V available for the BCI Competition III [
The data set was provided by the IDIAP Research Institute of Switzerland and undertakes the multiclass motor imagery problem. This set was recorded by a Biosemi system using a cap with 32 integrated electrodes located at standard positions of the International 10-20 system as depicted in Figure
Image of the montage applying the 10-20 system convention.
This data set focuses on a benchmark to classify three mental tasks [
The precomputed sets provided only include the sensors C3, Cz, C4, CP1, CP2, P3, Pz, and P4 out of the available 32 and they are the result of several transformations of the raw data. In the first stage, the potentials recorded were spatially filtered by means of a surface Laplacian. After that, a Power Spectral Density (PSD) calculation for the frequency band between 8 and 30 Hz with a resolution of 2 Hz was performed. Being the sampling frequency 512 Hz and the records divided in windows of 1 s with an additional rate of 32 samples, an overlapping of 93.75% between windows is defined.
The computational burden of this processing can be calculated as the product of 12 different features (or different frequencies bands) per sensor by 8 channels, involving a total of 96 features per sample, yielding 49,152 features per minute.
To facilitate the understanding of the results presented in this paper, Table
Channel and frequency associated with each feature in the input vector.
Channel | Frequency (Hz) | |||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
8 | 10 | 12 | 14 | 16 | 18 | 20 | 22 | 24 | 26 | 28 | 30 | |
C3 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 |
Cz | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 |
C4 | 25 | 26 | 27 | 28 | 29 | 30 | 31 | 32 | 33 | 34 | 35 | 36 |
CP1 | 37 | 38 | 39 | 40 | 41 | 42 | 43 | 44 | 45 | 46 | 47 | 48 |
CP2 | 49 | 50 | 51 | 52 | 53 | 54 | 55 | 56 | 57 | 58 | 59 | 60 |
P3 | 61 | 62 | 63 | 64 | 65 | 66 | 67 | 68 | 69 | 70 | 71 | 72 |
Pz | 73 | 74 | 75 | 76 | 77 | 78 | 79 | 80 | 81 | 82 | 83 | 84 |
P4 | 85 | 86 | 87 | 88 | 89 | 90 | 91 | 92 | 93 | 94 | 95 | 96 |
Out of the four available BCI Competition data sets per user, there are three learning data sets and a final one for testing. The learning sets are used to calculate the number of features selected by each one of the models, while the additional test session is only used at a later stage (Section
For the purpose of reducing the size of the features vector, a new methodology has been developed. Initially, the size of the features vector is the result of multiplying the number of channels used in the analysis by the number of frequencies considered in the PSD calculation. The classification method used is based on the S-dFasArt architecture proposed by Cano-Izquierdo et al. [
Figure
Feature selection proposed methodology.
Then, all the scores are added according to each feature, allowing the creation of a feature classification from most to least discriminant nature. Using this ranking, a first selection of the candidate features to form the reduced vector is obtained.
Two methodologies, based on the training data sets, are evaluated to analyze the discriminant nature of each of the components of the feature vector: the first one is supported by applying classic statistics methods, while the second is based on the fuzzy logic interpretation of the classifier which gets created from the training data set.
The framework on this research can be defined as a classification problem of
When establishing the criteria to determine the discrimination capacity contribution of each of the features, the statistical entropy can be estimated as
Alternative criteria to show the discriminant information of each feature can be defined as
This expression has a maximum value of
An architecture to classify EEG data applying the same benchmark as proposed in the BCI Competition Data Set V is proposed by Cano-Izquierdo et al. [ First, a learning session is used to generate a rule set defining the model. After that, a different learning session is devoted to adjust the model parameters to be applied at the test stage. Then, a rule prune is performed where the rules contributing to a higher error than success rate are discarded. Finally, once all the possible combinations of the three learning sessions are used for stages 1 and 2, there are six models available. For each one, 16 vectors per second are processed. Then, due to the fact that a prediction is produced every half a second only, every model contributes to 8 possible alternatives. To choose among the 48 = 6 × 8 possible predictions, a voting strategy is used where the most frequent prediction is selected.
S-dFasArt classification process.
For the purpose of feature selection, the third stage of the model is replaced by an “intermediate” model, which is defined with only three rules (each one associated with one single class). To do this, the weights defining every rule are calculated as the mean of the weights predicting the same category. The S-dFasArt model allows each class to be interpreted as a rule whose transference function is determined by the weights associated with fuzzy sets. Moreover, the rule associated with the
Also, it is assumed that the discriminant character of each feature will be linked to the relationship between its associated fuzzy sets for two classes. If these fuzzy sets are very similar, the feature will not be very discriminant. If the fuzzy sets are clearly different, the discriminant character of the feature will increase.
For each feature, the discriminant character is obtained by comparing the corresponding fuzzy sets for two rules
A value of
To determine the minimum number of features that can be part of the system while maintaining the output accuracy, the criteria based on the accumulated scores with regard to the total punctuation are presented. The scores are calculated by using both statistics method and fuzzy criteria. After that, the features are sorted in a descending order and the number of candidate features to be part of the model
The design parameter
This section summarizes the outcome of the application of the previous methodology and architecture to the BCI Competition III Data Set V database, addressing a three-class classification problem. First, the application of the statistics method is presented and the results for both Order and GMDH Selection are shown in different figures and tables. Then, the analogue information is shown for the methods based on fuzzy criteria. Section
Figure
To determine the most discriminant features, they have been ordered from higher to lower value of
Features, channels, and related frequencies.
|
||||||||||
---|---|---|---|---|---|---|---|---|---|---|
Feature | 38 | 2 | 14 | 25 | 26 | 31 | 27 | 3 | 8 | 50 |
Channel | CP1 | C3 | Cz | C4 | C4 | C4 | C4 | C3 | C3 | CP2 |
Freq/Hz | 10 | 10 | 10 | 8 | 10 | 20 | 12 | 12 | 22 | 10 |
|
||||||||||
|
||||||||||
|
||||||||||
Feature | 26 | 2 | 1 | 3 | 13 | 27 | 14 | 74 | 5 | 25 |
Channel | C4 | C3 | C3 | C3 | Cz | C4 | Cz | Pz | C3 | C4 |
Freq/Hz | 12 | 10 | 8 | 12 | 8 | 12 | 10 | 10 | 16 | 8 |
|
||||||||||
|
||||||||||
|
||||||||||
Feature | 39 | 3 | 1 | 2 | 4 | 31 | 96 | 30 | 92 | 35 |
Channel | CP1 | C3 | C3 | C3 | C3 | C4 | P4 | C4 | P4 | C4 |
Freq/Hz | 12 | 12 | 8 | 10 | 14 | 20 | 30 | 18 | 22 | 28 |
Relevance classification based on the score calculated from the discriminant nature of each feature.
The numbers of candidate features obtained after applying the 85% criteria for each of the three studied users results are
The results are presented in Table
Success rate (in %) as a function of the input vector dimension (
|
|
|
|
|
|
|
Average |
---|---|---|---|---|---|---|---|
|
|||||||
1 | 68.25 | 59.83 | 68.58 | 59.78 | 66.91 | 67.86 | 65.20 |
2 | 86.24 | 79.06 | 87.72 | 73.48 | 79.03 | 75.17 |
|
3 | 85.96 | 78.40 | 85.26 | 67.66 | 79.41 | 69.15 | 77.64 |
4 | 84.84 | 79.15 | 87.08 | 66.80 | 81.94 | 65.17 | 77.50 |
5 | 77.05 | 78.40 | 79.74 | 62.44 | 71.51 | 64.71 | 72.31 |
6 | 74.52 | 77.13 | 79.40 | 54.30 | 73.62 | 66.40 | 70.90 |
7 | 78.17 | 72.18 | 80.13 | 76.63 | 76.90 | 74.00 | 76.34 |
8 | 77.66 | 76.87 | 80.77 | 63.88 | 77.10 | 78.44 | 75.79 |
9 | 80.10 | 77.53 | 81.25 | 72.65 | 76.84 | 77.69 | 77.68 |
|
|||||||
|
|||||||
1 | 44.82 | 62.56 | 57.60 | 46.00 | 59.09 | 49.91 | 53.33 |
2 | 62.56 | 71.44 | 72.90 | 67.34 | 78.91 | 67.63 | 70.13 |
3 | 68.69 | 75.43 | 72.00 | 68.52 | 75.98 | 63.82 |
|
4 | 61.18 | 65.08 | 78.25 | 65.41 | 64.53 | 63.25 | 66.28 |
5 | 67.40 | 67.33 | 71.34 | 55.88 | 69.97 | 64.11 | 66.01 |
6 | 62.96 | 67.45 | 72.44 | 65.67 | 73.76 | 63.77 | 67.68 |
7 | 61.38 | 68.37 | 65.90 | 58.44 | 64.61 | 58.52 | 62.87 |
8 | 62.13 | 61.34 | 69.59 | 53.74 | 67.88 | 63.62 | 63.05 |
9 | 62.76 | 64.58 | 67.91 | 56.94 | 63.28 | 60.54 | 62.67 |
10 | 62.44 | 68.14 | 67.08 | 56.02 | 74.57 | 51.58 | 63.31 |
|
|||||||
|
|||||||
1 | 38.60 | 28.27 | 00.00 | 42.73 | 38.61 | 34.55 | 30.46 |
2 | 42.12 | 54.23 | 45.55 | 45.12 | 50.85 | 47.37 | 47.54 |
3 | 42.41 | 38.00 | 51.83 | 45.94 | 49.15 | 41.85 | 44.86 |
4 | 40.12 | 52.75 | 49.10 | 47.90 | 50.38 | 48.39 | 48.11 |
5 | 41.45 | 54.91 | 50.29 | 49.45 | 50.47 | 48.60 | 49.20 |
6 | 43.43 | 50.55 | 48.31 | 49.04 | 55.75 | 56.92 |
|
7 | 42.06 | 50.32 | 47.30 | 47.37 | 54.47 | 49.53 | 48.51 |
8 | 40.64 | 48.31 | 44.39 | 45.74 | 50.15 | 49.36 | 46.43 |
9 | 42.35 | 53.27 | 44.88 | 44.16 | 49.85 | 47.78 | 47.05 |
10 | 40.55 | 58.97 | 43.08 | 44.92 | 53.65 | 50.00 | 48.53 |
11 | 46.05 | 51.29 | 44.27 | 41.91 | 52.95 | 51.66 | 48.02 |
12 | 47.91 | 54.96 | 44.59 | 50.03 | 48.10 | 54.47 | 50.01 |
13 | 46.69 | 50.73 | 46.48 | 45.71 | 50.64 | 52.25 | 48.75 |
14 | 44.27 | 55.34 | 50.73 | 48.16 | 44.71 | 52.69 | 49.32 |
15 | 42.59 | 47.75 | 44.30 | 50.82 | 49.47 | 51.52 | 47.74 |
Following the same criteria,
Table
Features and RC values for the models calculated for the different users based on GMDH selection for statistics method scored data.
Features | RC |
---|---|
|
|
|
67.96 |
|
80.12 |
|
|
|
|
|
|
|
57.22 |
|
70.13 |
|
|
|
|
|
|
|
49.58 |
|
51.70 |
|
|
Figure
If the features are sorted from the highest to lowest value of
Relevance classification based on the score calculated from the discriminant nature of each feature.
When applying the 85% criteria on the value of
The best ten channels and the frequency value attached to them for every user are provided in Table
Features, channels, and related frequencies.
|
||||||||||
---|---|---|---|---|---|---|---|---|---|---|
Feature | 26 | 27 | 25 | 38 | 2 | 3 | 61 | 39 | 62 | 14 |
Channel | C4 | C4 | C4 | CP1 | C3 | C3 | P3 | CP1 | P3 | Cz |
Freq/Hz | 10 | 12 | 8 | 10 | 10 | 12 | 8 | 12 | 10 | 10 |
|
||||||||||
|
||||||||||
|
||||||||||
Feature | 2 | 26 | 1 | 25 | 74 | 27 | 3 | 14 | 73 | 13 |
Channel | C3 | C4 | C3 | C4 | Pz | C4 | C3 | Cz | Pz | Cz |
Freq/Hz | 10 | 10 | 8 | 8 | 10 | 12 | 12 | 10 | 8 | 8 |
|
||||||||||
|
||||||||||
|
||||||||||
Feature | 3 | 74 | 39 | 4 | 73 | 27 | 25 | 1 | 86 | 26 |
Channel | C3 | Pz | CP1 | C3 | Pz | C4 | C4 | C3 | P4 | C4 |
Freq/Hz | 12 | 10 | 12 | 14 | 8 | 12 | 8 | 8 | 10 | 10 |
Table
Success rate (in %) as a function of the input vector dimension (
|
|
|
|
|
|
|
Average |
---|---|---|---|---|---|---|---|
|
|||||||
1 | 49.78 | 55.62 | 39.27 | 40.54 | 54.21 | 58.60 | 49.67 |
2 | 56.98 | 66.04 | 60.17 | 55.39 | 70.28 | 71.65 | 63.42 |
3 | 65.30 | 60.17 | 65.98 | 59.32 | 67.40 | 59.46 | 62.94 |
4 | 78.08 | 73.73 | 78.78 | 65.14 | 74.02 | 70.53 | 73.38 |
5 | 78.76 | 73.44 | 80.47 | 78.35 | 77.45 | 69.58 | 76.34 |
6 | 79.71 | 70.19 | 80.63 | 76.86 | 77.85 | 65.97 | 75.20 |
7 | 79.23 | 78.51 | 78.62 | 58.97 | 79.98 | 68.15 | 73.91 |
8 | 79.07 | 77.79 | 79.48 | 71.62 | 76.50 | 69.47 | 75.66 |
9 | 80.30 | 79.00 | 80.04 | 66.37 | 81.08 | 81.07 |
|
|
|||||||
|
|||||||
1 | 34.48 | 58.33 | 65.87 | 66.07 | 60.94 | 57.63 | 57.22 |
2 | 62.56 | 71.44 | 72.90 | 67.34 | 78.91 | 67.63 | 70.13 |
3 | 68.69 | 75.43 | 72.00 | 68.52 | 75.98 | 63.82 |
|
4 | 61.43 | 63.31 | 70.56 | 59.50 | 69.53 | 62.33 | 64.44 |
5 | 61.46 | 70.05 | 67.17 | 59.65 | 63.54 | 53.74 | 62.60 |
6 | 58.06 | 69.70 | 72.87 | 58.41 | 61.83 | 62.07 | 63.83 |
7 | 68.03 | 65.74 | 65.84 | 58.38 | 45.60 | 61.95 | 60.92 |
8 | 65.01 | 62.24 | 50.60 | 43.98 | 70.95 | 58.67 | 58.58 |
9 | 68.15 | 66.20 | 62.41 | 48.10 | 66.30 | 58.64 | 61.63 |
10 | 65.67 | 66.41 | 65.15 | 63.59 | 66.26 | 56.45 | 63.92 |
11 | 52.33 | 67.36 | 68.63 | 58.27 | 56.57 | 55.90 | 59.84 |
|
|||||||
|
|||||||
1 | 42.50 | 54.61 | 34.65 | 46.41 | 54.47 | 45.59 | 46.37 |
2 | 41.28 | 36.01 | 40.17 | 46.82 | 49.97 | 46.82 | 43.51 |
3 | 43.87 | 48.60 | 44.22 | 46.90 | 48.01 | 46.06 | 46.28 |
4 | 44.39 | 53.21 | 41.95 | 48.83 | 51.26 | 48.48 | 48.02 |
5 | 45.17 | 54.38 | 42.88 | 44.89 | 53.91 | 49.45 |
|
6 | 42.67 | 51.26 | 43.90 | 47.14 | 56.89 | 47.20 | 48.18 |
7 | 39.85 | 56.75 | 41.10 | 42.73 | 48.57 | 44.57 | 45.60 |
8 | 41.16 | 54.56 | 43.69 | 46.73 | 45.39 | 45.59 | 46.19 |
9 | 42.03 | 50.29 | 40.38 | 43.55 | 46.47 | 45.85 | 44.76 |
10 | 41.31 | 50.32 | 38.75 | 43.43 | 44.10 | 38.70 | 42.77 |
11 | 40.32 | 45.97 | 38.49 | 48.16 | 44.71 | 44.60 | 43.71 |
12 | 41.19 | 50.18 | 44.62 | 41.44 | 42.79 | 45.24 | 44.24 |
13 | 42.01 | 44.80 | 36.31 | 42.03 | 44.92 | 41.65 | 41.96 |
14 | 41.69 | 47.81 | 41.69 | 47.96 | 45.74 | 41.44 | 44.39 |
15 | 41.42 | 51.90 | 39.22 | 46.99 | 42.41 | 43.05 | 44.17 |
16 | 40.23 | 41.12 | 37.65 | 43.49 | 39.08 | 43.87 | 40.91 |
From them, the input vector for
Analogously to the process followed for the statistic criteria, the GMDH method will be used with the purpose of selecting a model from a candidate feature set. In Table
Features and RC values for the models calculated for the different users based on GMDH selection for fuzzy criteria scored data.
Features | RC |
---|---|
|
|
|
67.96 |
|
80.12 |
|
80.77 |
|
|
|
|
|
|
|
57.22 |
|
70.13 |
|
70.89 |
|
|
|
|
|
|
|
49.58 |
|
51.70 |
|
|
It is fundamental to outline that the test set of the BCI Competition is first used in the calculations required to obtain the results presented in this section. In previous sections, only the learning session data sets are applied. In order to check the efficiency of the proposed methodology, a final stage has been performed following the method developed in [
Results for the test session.
Selection method | Model | Success rate | Number | % |
---|---|---|---|---|
|
||||
None |
|
87.21 | 96 | 100.00 |
Statistic + Order |
|
85.39 | 2 | 2.08 |
Statistic + GMDH |
|
87.64 | 3 | 3.13 |
Fuzzy + Order |
|
|
9 | 9.38 |
Fuzzy + GMDH |
|
89.50 | 4 | 4.17 |
|
||||
|
||||
None |
|
82.26 | 96 | 100.00 |
Statistic + Order |
|
81.80 | 3 | 3.13 |
Statistic + GMDH |
|
81.57 | 3 | 3.13 |
Fuzzy + Order |
|
81.80 | 3 | 3.13 |
Fuzzy + GMDH |
|
|
4 | 4.17 |
|
||||
|
||||
None |
|
58.72 | 96 | 100.00 |
Statistic + Order |
|
57.57 | 6 | 6.25 |
Statistic + GMDH |
|
|
3 | 3.13 |
Fuzzy + Order |
|
52.52 | 5 | 5.21 |
Fuzzy + GMDH |
|
57.80 | 3 | 3.13 |
|
||||
Average | ||||
None | 76.06 | 96.00 | 100 | |
Statistic + Order | 74.92 | 3.67 | 3.82 | |
Statistic + GMDH | 76.2 | 3.00 | 3.13 | |
Fuzzy + Order | 74.76 | 5.67 | 5.91 | |
Fuzzy + GMDH |
|
3.67 | 3.82 |
The most striking result to emerge from the data is that a reduction from a total of 96 to a range between 3 and 9 features is achieved. Interestingly, the classification success rate is maintained or even slightly improved while reducing the number of features.
Aler et al. [
Similarly, another approach for feature selection is presented in [
A comparison among the classification success rate of the BCI Competition Winner, the results presented in [
Research classification success rate comparison.
Selection method |
|
Feat |
|
Feat |
|
Feat | Av. |
|
|||||||
BCI Competition Winner | 79.60 | 96 | 70.31 | 96 | 56.02 | 96 | 68.65 |
MDLA [ |
79.68 | 9 | 66.82 | 17 | 54.59 | 1 | 67.03 |
SVM with evolved spatial + frequency-selection filters [ |
78.14 | 32 | 71.33 | 16 | 59.07 | 40 | 69.58 |
EEG Mapping [ |
85.71 | 8 | 73.80 | 8 | 64.28 | 8 | 74.60 |
Statistic + GMDH |
|
|
|
|
|
|
|
It is apparent from Table
The correlation between the selected features and the users has been tested too. However, a set of common features cannot be generalized. The results show how
Turning now to the channel position associated with the selected features (Figure
Table
Selected channels and frequencies for the Fuzzy + GMDH selection method.
Frequency (Hz) | ||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
8 | 10 | 12 | 14 | 16 | 18 | 20 | 22 | 24 | 26 | 28 | 30 | |
|
||||||||||||
C3 | · |
|
· | · | · | · | · | · | · | · | · | · |
Cz | · | · | · | · | · | · | · | · | · | · | · | · |
C4 | · | · |
|
· | · | · | · | · | · | · | · | · |
CP1 | · |
|
· | · | · | · | · | · | · | · | · | · |
CP2 | · | · | · | · | · | · | · | · | · | · | · | · |
P3 | · |
|
· | · | · | · | · | · | · | · | · | · |
Pz | · | · | · | · | · | · | · | · | · | · | · | · |
P4 | · | · | · | · | · | · | · | · | · | · | · | · |
|
||||||||||||
|
||||||||||||
C3 | · |
|
· | · | · | · | · | · | · | · | · | · |
Cz | · | · | · | · | · | · | · | · | · | · | · | |
C4 | · |
|
|
· | · | · | · | · | · | · | · | · |
CP1 | · | · | · | · | · | · | · | · | · | · | · | · |
CP2 | · | · | · | · | · | · | · | · | · | · | · | · |
P3 |
|
· | · | · | · | · | · | · | · | · | · | · |
Pz | · | · | · | · | · | · | · | · | · | · | · | · |
P4 | · | · | · | · | · | · | · | · | · | · | · | · |
|
||||||||||||
|
||||||||||||
C3 | · | · |
|
|
· | · | · | · | · | · | · | · |
Cz | · | · | · | · | · | · | · | · | · | · | · | · |
C4 | · | · | · | · | · | · | · | · | · | · | · | · |
CP1 | · | · | · | · | · | · | · | · | · | · | · | · |
CP2 |
|
· | · | · | · | · | · | · | · | · | · | · |
P3 | · | · | · | · | · | · | · | · | · | · | · | · |
Pz | · | · | · | · | · | · | · | · | · | · | · | · |
P4 | · | · | · | · | · | · | · | · | · | · | · | · |
This sensor selection matches neurophysiological literature as in [
Also, the data set comprises a status which is not related to motor imagery, like it is imagining words beginning with the same random letter. This one could activate other areas of the brain and cause features not included in the previous research to appear as highly discriminant in our model.
The processing cost per feature added to the model has also been calculated for each subject. At the preprocessing stage, and due to the calculations performed by the Welch periodogram PSD function, the time consumption is linear with the number of features and everyone’s preprocessing cost is 1.04% of the total. The neurofuzzy algorithm explained in this paper requires an increase of 9.21% of the processing time per feature during the model generation (learning and rule prune), which is very significant considering that six models are generated for each user. A final 7.53% increase at the test stage for every feature added to the model is also required.
Table
Research classification success rate comparison.
Selection method |
|
Feat |
|
Feat |
|
Feat | Av. |
|
|||||||
BCI Competition Winner | 79.60 | 96 | 70.31 | 96 | 56.02 | 96 | 68.65 |
MDLA [ |
79.68 | 9 | 66.82 | 17 | 54.59 | 1 | 67.03 |
SVM with evolved spatial + frequency-selection filters [ |
78.14 | 32 | 71.33 | 16 | 59.07 | 40 | 69.58 |
EEG Mapping [ |
85.71 | 8 | 73.80 | 8 | 64.28 | 8 | 74.60 |
Statistic + GMDH |
|
|
|
|
|
|
|
|
|||||||
Unified model | 83.56 | 5 | 78.34 | 5 | 56.42 | 5 | 72.77 |
As can be found, the accuracy is slightly lower than that in the user specific models, but the reduction is only a 3.43% and the results are only improved by those shown in EEG Mapping [
A further investigation on this field should be carried out across a larger population to determine if a reduced set of common features across the users can be found as performed by Fazli et al. [
The most obvious finding to emerge from this study is a way of drastically reducing the number of features required on the processing of the BCI systems while maintaining and even improving their classification success rate. This approach, being a three-status paradigm where only two of them are motor imagery related, has not been commonly undertaken by the literature.
The results of this investigation show that a 96% reduction of the required number of features (from 96 to 4) for a selection method based on Fuzzy and GMDH algorithms can be achieved. This translates into important time saving in computational burden when the analysis of the time consumption is performed over this simplified model.
Moreover, the methodology proposed presents a native support to multiclass problems. Most of the research papers focus on reducing channels in two tasks motor imagery paradigms. Therefore, two-class classification algorithms are an excellent tool to address the problem yielding good results in terms of the calculation time and accuracy. However, when increasing the number of classes within the problem, feature selection methods based on algorithms such as CSF, FDA, SVM, and FC require a review of the entire system and the inclusions of decision trees. In addition, the calculations need to be repeated several times in two-class space combinations, increasing the processing time and power consumption before reaching an outcome.
In contrast, the use of S-dFasArt does not require any further tuning when increasing the number of classes and the processing time remains the same due to the fact that no new calculations are being required.
It has also been shown how the user and the features selected present an important correlation. As previous studies have reported, it has been found that the
Further experimental investigations are needed to estimate the smallest number of common features required for the exercise presented in this paper across a larger population. An important practical implication of this would be the manufacturing of low-cost headsets with a small number of sensors. Also, the processing should be quicker as the preprocessing stage and the classification algorithm would only perform calculations on a very small set of the sampled data. Therefore, the design of devices including a reduced number of sensors could be possible. This would allow the EEG systems to be more user friendly by drastically reducing the setup time. Also, more appealing headsets compared with the current cap system could be manufactured.
In summary, it has been demonstrated that the analysis of only a few frequency bands is required. This allows an important saving in computation time and power consumption as well, which is beneficial when integrating the system, due to the fact that less processing power and memory resources are being required. The aforementioned benefits can be critical when designing applications where the available times to provide them with an output or the hardware platform are limited, for example, in applications for mobile devices.
As a consequence of the reduction in the hardware, the creation of an affordable mass market mobile system based on EEG would be possible.
The authors declare that there is no conflict of interests regarding the publication of this paper.