Citation Context

... ) . (4) The log likelihood function given in (4) is maximized using the Expectation-Maximization algorithm [14]. The parameter K can be determined experimentally using Akaike’s information criterion =-=[15]-=-. It is often used to determine an appropriate number of mixture components when the number of components is unspecified. Akaike information is the negative log-likelihood for the data with a penalty ...

Citation Context

...e to form two types of geometrical facial features as described below. The CK+ database provides the locations of 68 points on the face at each frame, which are tracked using Active Appearance Models =-=[10, 13]-=-.There are 123 subjects and 327 emotion labeled image sequences in the CK+ database belonging to the six 1385978-1-4673-0046-9/12/$26.00 ©2012 IEEE ICASSP 2012 basic emotions. Image sequences start wi...

Citation Context

... the valence, activation and dominance coordinates [6]. Many studies have been published on affect recognition from facial expressions in the last decade, which are summarized in recent survey papers =-=[7, 6, 8, 9]-=-. Most of these methods use two dimensional spatio-temporal facial features, which are fed to a pattern recognition algorithm. Facial features can be categorized as geometrical features and appearance...

Citation Context

... the valence, activation and dominance coordinates [6]. Many studies have been published on affect recognition from facial expressions in the last decade, which are summarized in recent survey papers =-=[7, 6, 8, 9]-=-. Most of these methods use two dimensional spatio-temporal facial features, which are fed to a pattern recognition algorithm. Facial features can be categorized as geometrical features and appearance...

Citation Context

...racted a lot of interest in recent years. It is envisioned that the capability of recognizing human emotions will be a part of manmachine interactions and ubiquitous computing scenarios in the future =-=[1]-=-. Facial expression recognition has many other application areas including security [2], driver safety [3] and health-care [4]. 1.1. Previous Work A major problem in classifying facial expressions is ...

Citation Context

...identity-specific component can be eliminated by subtracting the features obtained from a neutral facial expression of that person from the current frame, which may be the first frame of a video clip =-=[10]-=- as in CK+ database. However, neutral face information of that person may not always be available. In that case, researchers generally average the features of a certain number of images in the video c...

Citation Context

... manmachine interactions and ubiquitous computing scenarios in the future [1]. Facial expression recognition has many other application areas including security [2], driver safety [3] and health-care =-=[4]-=-. 1.1. Previous Work A major problem in classifying facial expressions is defining the emotion classes. A solution to this has been suggested by Ekman [5], who specified a universally displayed set of...

Citation Context

..., sadness and surprise. Another solution, which has gained popularity recently is to use dimensional and continuous labeling of the affective cues in the valence, activation and dominance coordinates =-=[6]-=-. Many studies have been published on affect recognition from facial expressions in the last decade, which are summarized in recent survey papers [7, 6, 8, 9]. Most of these methods use two dimensiona...

Citation Context

... the valence, activation and dominance coordinates [6]. Many studies have been published on affect recognition from facial expressions in the last decade, which are summarized in recent survey papers =-=[7, 6, 8, 9]-=-. Most of these methods use two dimensional spatio-temporal facial features, which are fed to a pattern recognition algorithm. Facial features can be categorized as geometrical features and appearance...

Citation Context

...nizing human emotions will be a part of manmachine interactions and ubiquitous computing scenarios in the future [1]. Facial expression recognition has many other application areas including security =-=[2]-=-, driver safety [3] and health-care [4]. 1.1. Previous Work A major problem in classifying facial expressions is defining the emotion classes. A solution to this has been suggested by Ekman [5], who s...

Citation Context

... expression is not available. ii) We also experimentally evaluate two different geometric features, which we call the coordinate based features (CBF) [10] and distance and angle based features (DABF) =-=[12]-=-. CBF features have been observed to give higher emotion recognition rates on the CK+ database. In Section 2, a more detailed description of the used geometrical features are described. In Section 3, ...

Citation Context

...ns will be a part of manmachine interactions and ubiquitous computing scenarios in the future [1]. Facial expression recognition has many other application areas including security [2], driver safety =-=[3]-=- and health-care [4]. 1.1. Previous Work A major problem in classifying facial expressions is defining the emotion classes. A solution to this has been suggested by Ekman [5], who specified a universa...

Citation Context

...curity [2], driver safety [3] and health-care [4]. 1.1. Previous Work A major problem in classifying facial expressions is defining the emotion classes. A solution to this has been suggested by Ekman =-=[5]-=-, who specified a universally displayed set of six emotions: anger, happiness, disgust, fear, sadness and surprise. Another solution, which has gained popularity recently is to use dimensional and con...