EEG-based trial-by-trial texture classification during active touch

Abstract

Trial-by-trial texture classification analysis and identifying salient texture related EEG features during active touch that are minimally influenced by movement type and frequency conditions are the main contributions of this work. A total of twelve healthy subjects were recruited. Each subject was instructed to use the fingertip of their dominant hand's index finger to rub or tap three textured surfaces (smooth flat, medium rough, and rough) with three levels of movement frequency (approximately 2, 1 and 0.5 Hz). EEG and force data were collected synchronously during each touch condition. A systematic feature selection process was performed to select temporal and spectral EEG features that contribute to texture classification but have low contribution towards movement type and frequency classification. A tenfold cross validation was used to train two 3-class (each for texture and movement frequency classification) and a 2-class (movement type) Support Vector Machine classifiers. Our results showed that the total power in the mu (8-15 Hz) and beta (16-30 Hz) frequency bands showed high accuracy in discriminating among textures with different levels of roughness (average accuracy > 84%) but lower contribution towards movement type (average accuracy < 65%) and frequency (average accuracy < 58%) classification.

Figure 1

The mean and standard deviation of the classification accuracy of each classification problem (texture, movement type and frequency) calculated at each feature set and averaged across all the participants and all conditions and averaged over the three EEG channels (C1, C3 and C5). Each group of feature sets is highlighted by different background color. The selected set of features are marked by the blue and red arrows.

Figure 2

(A) The mean and standard deviation of the accuracy, sensitivity of each class (flat, medium-rough and rough surfaces) calculated at the electrode channel with the highest accuracy value and averaged across all participants for each condition. S1: flat, S2: medium-rough and S3 is the rough surface. (B) Scalp topography for the average accuracy values at each electrode location across all participants, where each condition is as follows, I: rub at fast movement frequency, II: rub at medium movement frequency, III: rub at slow movement frequency, IV: tap at fast movement frequency, V: tap at medium movement frequency and VI is tap at slow movement frequency.

Figure 3

(A) The mean and standard deviation of the accuracy, sensitivity of each class (rub and tap) calculated at the electrode channel with the highest accuracy value and averaged across all participants for each condition. S1: flat, S2: medium-rough and S3 is the rough surface. (B) Scalp topography for the average accuracy values at each electrode location across all participants, where each condition is as follows, I: flat surface, fast movement frequency, II: med-rough surface, fast movement frequency, III: rough surface, fast movement frequency IV: flat surface at medium movement frequency V: med-rough surface at medium movement frequency , VI: rough surface at medium movement frequency , VII: flat surface at slow movement frequency, VIII: medium-rough surface at slow movement frequency and IX: is rough surface at slow movement frequency.

Figure 4

(A) The mean and standard deviation of the accuracy, sensitivity of each class (fast, medium movement frequency and slow) calculated at the electrode channel with the highest accuracy value and averaged across all participants for each condition. S1: flat, S2: medium-rough and S3 is the rough surface. (B) Scalp topography for the average accuracy values at each electrode location across all participants, where each condition is as follows, I: flat surface, rub, II: med-rough surface, rub, III: rough surface, rub IV: flat surface, tap V: med-rough surface, tap, VI: rough surface, tap.

Figure 5

The power spectral density of the medium rough and rough surfaces respectively.

Figure 6

Experimental Setup. (A) The participant is tapping the texture mounted on the force transducer. (B) Schematic diagram showing the tap and rub movement conditions.

Figure 7

(A) The normal force profile, in blue, and the result of applying a first order derivative filter in red. The green arrows show the local peaks that mark the beginning and end of each trial. (B) An illustration showing the 18 conditions, and the trials within each condition.