Subject-independent decoding of affective states using functional near-infrared spectroscopy

Figures

• 

  Fig 1. Channel configuration. Red circles represent sources; blue circles represent the detectors and dotted lines the channels. Colors show channel positions according to the International 10–20 system.

• 

  Fig 2. Visual stimuli order in reactive and active trials. The order of trials into the block is random, but always alternating neutral trials with positive and negative affect trials.

• 

  Fig 3. (a) Feature extraction procedure, showing steps from the whole raw signal until the creation of a feature table per subject. (b) Machine learning procedure: Initially, we performed an LDA-based feature selection using the first experimental sample (49 subjects). Then, based on these selected features, we created training and test sets for the LDA classifier using the first (49 subjects) and second (20 subjects) experimental samples, respectively.

• 

  Fig 4. Box plots showing the results using the LDA classifier and different feature subsets. (a) presents the “Positive vs. Negative”, (b) the “Positive vs. Neutral”, and (c) the “Negative vs. Neutral” comparisons for reactive elicitation block, while (d-f) follows the same order for active elicitation block. Black dots present the means, red lines the medians, red crosses the outliers and black asterisks the statistical difference for chance level (p<0.05).

• 

  Fig 5. Weights assigned by the LDA classifier for each feature during the (a) "Positive vs. Negative" comparisons using 20% of features and (b) "Negative vs. Neutral" comparisons using 40% of features. In (c), the weights for each feature during the "Positive vs. Neutral" comparisons using 5% of features. Hottest colors indicate positive weights while cooler colors indicate negative weights. Channels filled with gray dots were not used during the test.

• 

  Fig 6. Averaged time series and features used as inputs to the LDA classifier during “Positive vs. Neutral” comparison in the active task. Data from Channel 6–7 is shown in (a-b), Channel 5–4 in (c-d) and Channel 4–4 in (e-f) for the training and test samples, respectively. Blue curves and boxes represent Neutral trials, and red curves and boxes Positive trials.