780 likes | 1k Views
Emotion Recognition on the Basis of Peripheral Physiological Data. Dipl.-Inf. David Hrabal Medizinische Psychologie Uniklinikum Ulm. Ulm, 17.12.2012 Doktorandenseminar. Emotion Recognition on the Basis of Peripheral Physiological Data. Emotion Recognition on the Basis of
E N D
Emotion Recognition on the Basis of Peripheral Physiological Data Dipl.-Inf. David Hrabal Medizinische Psychologie Uniklinikum Ulm Ulm, 17.12.2012 Doktorandenseminar
Emotion Recognition on the Basis of Peripheral Physiological Data
Emotion Recognition on the Basis of Peripheral Physiological Data • What is an emotion? • What is physiological data? • Common classification methods • New approach to emotion recognition on the basis of physiological data - 'Feature-Pair Based Emotion Identification'
What is an emotion? • There are many definitions for “Emotion”
What is an emotion? • There are many definitions for “Emotion” • It is known that emotions play a major role in: motivation, perception, cognition, coping, creativity, attention, planning, reasoning, learning, memory, decision making...
What is an emotion? • There are many definitions for “Emotion” • It is known that emotions play a major role in: motivation, perception, cognition, coping, creativity, attention, planning, reasoning, learning, memory, decision making... • Like many other procedures, it consists of: sensory input (perception), action (neurological processing), output (expression)
What is an emotion? • Neurophysiological action -> activation / inhibition -> changes in: heart rate, respiration frequency / intensity, temperature, muscle tension, brain wave changes
What is physiological data? Data that was recorded using biosensors:
What is physiological data? Data that was recorded using biosensors: • EMG (electromyography)
What is physiological data? Data that was recorded using biosensors: • EMG (electromyography) • SC (skin conductance)
What is physiological data? Data that was recorded using biosensors: • EMG (electromyography) • SC (skin conductance) • BVP (blood volume pulse)
What is physiological data? Data that was recorded using biosensors: • EMG (electromyography) • SC (skin conductance) • BVP (blood volume pulse) • Temperature • ECG (electrocardiography) • Respiration • EEG (electromyography) • ...
What is an emotion? • Neurophysiological action -> activation / inhibition -> changes in: heart rate, respiration frequency / intensity, temperature, muscle tension, brain wave changes Since it is possible to measure physiological changes with biosensors and emotion leads to physiological changes, it must be possible to measure emotion with biosensors!
Common Methods for Emotion Recognition Method 1: Machine Learning
Common Methods for Emotion Recognition Method 1: Machine Learning selected features as inputs for an automatic classification system, a classification rate is calculated
Common Methods for Emotion Recognition Method 1: Machine Learning selected features as inputs for an automatic classification system, a classification rate is calculated • From the physiological signal
Common Methods for Emotion Recognition Method 1: Machine Learning selected features as inputs for an automatic classification system, a classification rate is calculated • From the physiological signal • statistical features are calculated • (extracted)
Common Methods for Emotion Recognition Method 1: Machine Learning selected features as inputs for an automatic classification system, a classification rate is calculated • From the physiological signal • statistical features are calculated • (extracted) • The features are used as input • vector for a SVM or NN
Common Methods for Emotion Recognition Method 1: Machine Learning selected features as inputs for an automatic classification system, a classification rate is calculated • From the physiological signal • statistical features are calculated • (extracted) • The features are used as input • vector for a SVM or NN • The classification rate is the result • of a leave-one-out classification
Common Methods for Emotion Recognition Method 1: Machine Learning
Common Methods for Emotion Recognition Method 1: Machine Learning
Common Methods for Emotion Recognition Method 1: Machine Learning
Common Methods for Emotion Recognition Method 1: Machine Learning good results in leave-one-out classification transsituational classification isn't calculated or delivers bad results
Common Methods for Emotion Recognition Method 1: Machine Learning good results in leave-one-out classification transsituational classification isn't calculated or delivers bad results PROBLEM: the calculated features help specify this one recorded dataset but (in most cases) they are not transferable
Common Methods for Emotion Recognition Method 2: Basic Research
Common Methods for Emotion Recognition Method 2: Basic Research Individual channels are observed when trying to distignuish between emotional states like fear and sadness
Common Methods for Emotion Recognition Method 2: Basic Research Individual channels are observed when trying to distignuish between emotional states like fear and sadness fear sadness
Common Methods for Emotion Recognition Method 2: Basic Research - too ambiguous (mehrdeutig) : changes in a certain channel cannot be traced back to a certain emotion
Common Methods for Emotion Recognition Method 2: Basic Research - too ambiguous (mehrdeutig) : changes in a certain channel cannot be traced back to a certain emotion fear sadness
Common Methods for Emotion Recognition Method 2: Basic Research - too ambiguous (mehrdeutig) : changes in a certain channel cannot be traced back to a certain emotion - in this form not applicable for the basic emotion concept
Common Methods for Emotion Recognition Method 2: Basic Research - too ambiguous (mehrdeutig) : changes in a certain channel cannot be traced back to a certain emotion - in this form not applicable for the basic emotion concept - what about the dimensional concept? pleasure arousal
New Approach to Emotion Recognition from Physiological Data physiological channel EMG, SC or BVP
New Approach to Emotion Recognition from Physiological Data physiological channel change direction EMG, SC or BVP or
New Approach to Emotion Recognition from Physiological Data physiological channel change direction number of states EMG, SC or BVP or 2^1 = 2
New Approach to Emotion Recognition from Physiological Data physiological channel change direction number of states EMG, SC or BVP or 2^1 = 2 pleasure arousal
New Approach to Emotion Recognition from Physiological Data physiological channel change direction number of states EMG, SC or BVP or 2^1 = 2 EMG_corr EMG_zyg SC pleasure BVP arousal
New Approach to Emotion Recognition from Physiological Data physiological channel change direction number of states EMG, SC or BVP or 2^1 = 2 EMG_zyg pleasure arousal
New Approach to Emotion Recognition from Physiological Data physiological channel change direction number of states EMG, SC or BVP or 2^1 = 2 EMG_zyg pleasure arousal
New Approach to Emotion Recognition from Physiological Data physiological channel change direction number of states EMG, SC or BVP or 2^1 = 2 pleasure SC arousal
New Approach to Emotion Recognition from Physiological Data physiological channel change direction number of states EMG, SC or BVP or 2^1 = 2 pleasure SC arousal
New Approach to Emotion Recognition from Physiological Data physiological channel change direction number of states EMG, SC or BVP or 2^1 = 2 EMG_corr EMG_zyg pleasure SC BVP arousal
New Approach to Emotion Recognition from Physiological Data physiological channel change direction number of states EMG, SC or BVP or 2^1 = 2 EMG_corr EMG_zyg pleasure SC Each physiological parameter can affect one axis BVP arousal
New Approach to Emotion Recognition from Physiological Data physiological channel change direction number of states EMG, SC or BVP or 2^1 = 2
New Approach to Emotion Recognition from Physiological Data physiological channel change direction number of states EMG, SC or BVP or 2^1 = 2 4
New Approach to Emotion Recognition from Physiological Data physiological channel change direction number of states EMG, SC or BVP or 2^1 = 2 2^2 = 4
New Approach to Emotion Recognition from Physiological Data physiological channel change direction number of states EMG, SC or BVP or 2^1 = 2 EMG_zyg + X 2^2 = 4
New Approach to Emotion Recognition from Physiological Data physiological channel change direction number of states EMG, SC or BVP or 2^1 = 2 EMG_zyg + X , , or 2^2 = 4
New Approach to Emotion Recognition from Physiological Data physiological channel change direction number of states EMG, SC or BVP or 2^1 = 2 EMG_zyg + X , , or 2^2 = 4 EMG_zyg + X pleasure arousal