8. Affective Computing

Question 1

What is Affective Computing?

Answer 1

Development of computers/machines that recognize and express emotions

Interlaced with psychology, cognitive science, and computer science/electrical engineering

Question 2

Why do we want affective computing?

Answer 2

To enhance human-computer interaction by incorporating emotional understanding

Current interactions with computers are devoid of emotions

Question 3

How do humans modulate conversations based on emotions?

Answer 3

Humans adjust their tone and expressions depending on the other’s emotional state

Words can convey different meanings based on tone

Question 4

Give an example of how tone can change the meaning of a statement.

Answer 4

The response to ‘Are you mad at me?’ could mean ‘No, I’m not’ or ‘Yes I am but I don’t want to say’

Contextual understanding of tone is crucial

Question 5

Which types of computers may need affective abilities?

Answer 5

Assistive robots, such as elderly caretaker robots

These robots should regulate responses based on users’ emotions

Question 6

What are some ways a computer can recognize emotions?

Answer 6

• Facial expressions
• Tone analysis
• Pulse

Most computers are equipped with cameras and microphones

Question 7

How can a computer express emotions?

Answer 7

• Tone
• Emojis
• Capitalization
• Color

These methods help convey emotional context

Question 8

What is the James-Lange theory of emotion?

Answer 8

Emotions are physical processes; physiological responses occur first and are then labeled as emotions

Example: Encountering a bear triggers a fight or flight response

Question 9

What does the Cannon-Bard theory state?

Answer 9

Emotions originate in the brain; emotional experience and physiological changes occur simultaneously

Example: Seeing a bear and feeling fear at the same time

Question 10

What do most psychologists believe about the theories of emotion?

Answer 10

Both James-Lange and Cannon-Bard theories are correct to some extent

Pain can lead to anger/sadness (James-Lange) and thoughts can lead to emotions (Cannon-Bard)

Question 11

What are discrete emotions models?

Answer 11

Models that suggest there are only a certain number of basic emotions we can feel

Six basic emotions: anger, disgust, fear, happiness, sadness, surprise

Question 12

What are the two dimensions in the dimensional emotions model?

Answer 12

• Valence (pleasantness/unpleasantness)
• Arousal

Emotions are constructed from these dimensional components

Question 13

What is the third dimension of emotions that some psychologists propose?

Answer 13

Dominance, which reflects how much control one feels

Affective computing typically does not use this dimension

Question 14

What is the basic process of affect recognition?

Answer 14

• Provide input (image, audio, physiological symptoms)
• Input is read by algorithms
• Output is the emotion

Machine learning algorithms, specifically deep learning, are used for this process

Question 15

What is the first step in training a machine learning model for affect recognition?

Answer 15

Label all images with the emotion being displayed

This helps the model learn to recognize emotions from features extracted from images

Question 16

What should be done with images before feeding them to the algorithm?

Answer 16

Extract features from the images instead of feeding all the image data

This prevents overwhelming the algorithm

Question 17

What are the ways we can measure visual modality in affect recognition?

Answer 17

• Facial expression
• Body expression
• Eye gaze
• Pupil diameter
• Blinking behaviour

Facial expressions are considered the most intuitive indicators of affect.

Question 18

Who proposed that facial expressions are universal?

Answer 18

Charles Darwin

Darwin suggested that facial expressions are universally recognized but could not prove it.

Question 19

What experiment did Ekman conduct regarding facial expressions?

Answer 19

He photographed facial expressions of people from an isolated tribe.

The results showed their facial expressions were similar to those of non-isolated cultures.

Question 20

What are display rules in the context of facial expressions?

Answer 20

Cultural rules that dictate how emotions are expressed.

These rules can lead to amplification or suppression of emotions based on social context.

Question 21

What is the first step in the detection of facial expressions?

Answer 21

Face detection (and tracking)

It involves blocking out all other camera details to focus on the face.

Question 22

What is the most popular method for face detection?

Answer 22

Viola Jones

This method is widely used for real-time face detection.

Question 23

What are the two methods of feature extraction for facial expressions?

Answer 23

• Geometric features
• Appearance features

Question 24

What do geometric features involve in facial expression detection?

Answer 24

Using a mask of shapes to map facial points and measuring distances and velocity vectors.

This helps in tracking how facial expressions develop.

Question 25

What do appearance features focus on?

Answer 25

Applying image filters to detect changes in image texture. ## Footnote This includes looking for edges and creases like dimples or smile lines.

Question 26

What is affect classification?

Answer 26

Emotion classification ## Footnote It involves estimating emotions based on various modalities.

Question 27

What are the two channels of information in speech?

Answer 27

* Speech semantics * Speech intonation

Question 28

What is sentiment analysis?

Answer 28

The process of determining the sentiment behind words. ## Footnote It can be done using lexicon-based technology or statistical-learning techniques.

Question 29

How does a MEL spectrogram help in speech analysis?

Answer 29

It converts time domain signals into frequency/time domain to analyze intonation. ## Footnote It emphasizes low frequencies, aligning with human auditory perception.

Question 30

What does electromyography (EMG) measure?

Answer 30

Muscle response ## Footnote It detects muscle tension associated with emotions like fear.

Question 31

What does electrocardiography (ECG) provide information about?

Answer 31

Electrical signals of the heart ## Footnote It reflects the state of the autonomous nervous system impacted by emotions.

Question 32

What does electroencephalography (EEG) measure?

Answer 32

Electrical activity of the brain ## Footnote Higher signals indicate more thoughts or cognitive activity.

Question 33

What does skin conductance measure?

Answer 33

Galvanic skin response or electrodermal activity ## Footnote It indicates sweating levels, which increase under stress.

Question 34

What physiological response is associated with stress?

Answer 34

Faster respiration rate ## Footnote Stress often leads to accelerated breathing as part of the fight or flight response.

Question 35

What is multimodal fusion in emotion detection?

Answer 35

Connecting signals from multiple modalities to improve emotion recognition. ## Footnote This includes feature level and decision level fusion.

Question 36

What is feature level fusion?

Answer 36

Extracting features from each modality and combining them for machine learning analysis. ## Footnote This approach feeds all features into a machine learning model.

Question 37

What is decision level fusion?

Answer 37

Feeding features from each modality into specialized machine learning models and combining results. ## Footnote This helps in accurately determining emotions.

Question 38

What is hybrid fusion in emotion detection?

Answer 38

Combining feature level fusion with decision level fusion. ## Footnote It allows for a more comprehensive analysis of emotions.