https://schererstefan.net/assets/files/papers/1-s2.0-S0167639315000369-main.pdf

A review of depression and suicide risk assessment using speech analysis

We present a novel discriminative regression based approach for the Constrained Local Models (CLMs) framework, referred to as the Discriminative Response Map Fitting (DRMF) method, which shows impressive performance in the generic face fitting scenario. The motivation behind this approach is that, unlike the holistic texture based features used in the discriminative AAM approaches, the response map can be represented by a small set of parameters and these parameters can be very efficiently used for reconstructing unseen response maps. Furthermore, we show that by adopting very simple off-the-shelf regression techniques, it is possible to learn robust functions from response maps to the shape parameters updates. The experiments, conducted on Multi-PIE, XM2VTS and LFPW database, show that the proposed DRMF method outperforms state-of-the-art algorithms for the task of generic face fitting. Moreover, the DRMF method is computationally very efficient and is real-time capable. The current MATLAB implementation takes 1 second per image. To facilitate future comparisons, we release the MATLAB code and the pre-trained models for research purposes.

/pdf/robust-discriminative-response-map-fitting-with-constrained-zfoumjb9m3.pdf

Robust Discriminative Response Map Fitting with Constrained Local Models

The development of facial databases with an abundance of annotated facial data captured under unconstrained 'in-the-wild' conditions have made discriminative facial deformable models the de facto choice for generic facial landmark localization. Even though very good performance for the facial landmark localization has been shown by many recently proposed discriminative techniques, when it comes to the applications that require excellent accuracy, such as facial behaviour analysis and facial motion capture, the semi-automatic person-specific or even tedious manual tracking is still the preferred choice. One way to construct a person-specific model automatically is through incremental updating of the generic model. This paper deals with the problem of updating a discriminative facial deformable model, a problem that has not been thoroughly studied in the literature. In particular, we study for the first time, to the best of our knowledge, the strategies to update a discriminative model that is trained by a cascade of regressors. We propose very efficient strategies to update the model and we show that is possible to automatically construct robust discriminative person and imaging condition specific models 'in-the-wild' that outperform state-of-the-art generic face alignment strategies.

/pdf/incremental-face-alignment-in-the-wild-2vtxny75xk.pdf

Incremental Face Alignment in the Wild

Quality data recorded in varied realistic environments is vital for effective human face related research. Currently available datasets for human facial expression analysis have been generated in highly controlled lab environments. We present a new static facial expression database Static Facial Expressions in the Wild (SFEW) extracted from a temporal facial expressions database Acted Facial Expressions in the Wild (AFEW) [9], which we have extracted from movies. In the past, many robust methods have been reported in the literature. However, these methods have been experimented on different databases or using different protocols within the same databases. The lack of a standard protocol makes it difficult to compare systems and acts as a hindrance in the progress of the field. Therefore, we propose a person independent training and testing protocol for expression recognition as part of the BEFIT workshop. Further, we compare our dataset with the JAFFE and Multi-PIE datasets and provide baseline results.

/pdf/static-facial-expression-analysis-in-tough-conditions-data-12db8pndtq.pdf

Static facial expression analysis in tough conditions: Data, evaluation protocol and benchmark

We describe a very simple framework for deriving the most-well known optimization problems in Active Appearance Models (AAMs), and most importantly for providing efficient solutions. Our formulation results in two optimization problems for fast and exact AAM fitting, and one new algorithm which has the important advantage of being applicable to 3D. We show that the dominant cost for both forward and inverse algorithms is a few times mN which is the cost of projecting an image onto the appearance subspace. This makes both algorithms not only computationally realizable but also very attractive speed-wise for most current systems. Because exact AAM fitting is no longer computationally prohibitive, we trained AAMs in-the-wild with the goal of investigating whether AAMs benefit from such a training process. Our results show that although we did not use sophisticated shape priors, robust features or robust norms for improving performance, AAMs perform notably well and in some cases comparably with current state-of-the-art methods. We provide Matlab source code for training, fitting and reproducing the results presented in this paper at http://ibug.doc.ic.ac.uk/resources.

https://ibug.doc.ic.ac.uk/media/uploads/documents/tzimiro_pantic_iccv2013.pdf

Optimization Problems for Fast AAM Fitting in-the-Wild

http://users.cecs.anu.edu.au/~roland/Publications/SaragihGoecke_PatternRecognition2009.pdf

Learning AAM fitting through simulation

This paper is motivated by Ellgring's work in non-verbal communication in depression to measure and compare the levels of facial activity, before and after treatment, of endogenous and neurotic depressives. Similar to that work, we loosely associate the measurements with Action Units (AU) groups from the Facial Action Coding System (FACS). However, we use the neologism Region Units (RU) to describe regions of the face that encapsulate AUs. In contrast to Ell-gring's approach, we automatically generate the measurements and provide both prototypical expression recognition and RU-specific activity measurements. Latency between expressions is also measured and the system is conducive to comparison across groups and individual subjects. By using Active Appearance Models (AAM) to locate the fiduciary facial points, and MultiBoost to classify prototypical expressions and the RUs, we can provide a simple, objective, flexible and cost-effective means of automatically measuring facial activity.

/pdf/an-approach-for-automatically-measuring-facial-activity-in-41pegog81v.pdf

An approach for automatically measuring facial activity in depressed subjects

This paper describes ongoing work towards building a multimodal computer system capable of sensing the affective state of a user. Two major problem areas exist in the affective communication research. Firstly, affective states are defined and described in an inconsistent way. Secondly, the type of training data commonly used gives an oversimplified picture of affective expression. Most studies ignore the dynamic, versatile and personalised nature of affective expression and the influence that social setting, context and culture have on its rules of display. We present a novel approach to affective sensing, using a generic model of affective communication and a set of ontologies to assist in the analysis of concepts and to enhance the recognition process. Whilst the scope of the ontology provides for a full range of multimodal sensing, this paper focuses on spoken language and facial expressions as examples.

/pdf/towards-affective-sensing-h2g8pd4jr6.pdf

Towards affective sensing

Interest in emotion detection is increasing significantly For research and development in the field of Affective Computing, in smart environments, but also for reliable non-lab medical and psychological studies or human performance monitoring, robust technologies are needed for detecting evidence of emotions in persons under everyday conditions This paper reports on evaluation studies of the EREC-II sensor system for acquisition of emotion-related physiological parameters The system has been developed with a focus on easy handling, robustness, and reliability Two sets of studies have been performed covering 4 different application fields: medical, human performance in sports, driver assistance, and multimodal affect sensing Results show that the different application fields pose different requirements mainly on the user interface, while the hardware for sensing and processing the data proved to be in an acceptable state for use in different research domains

/pdf/erec-ii-in-use-studies-on-usability-and-suitability-of-a-4ys2ckqod5.pdf

EREC-II in use: studies on usability and suitability of a sensor system for affect detection and human performance monitoring

This paper describes some of the issues faced by typical emotion recognition systems and the need to be able to deal with emotions in a natural setting. Studies tend to ignore the dynamic, versatile and personalised nature of affective expression and the influence that social setting, context and culture have on its rules of display. Affective cues can be present in multiple modalities and they can manifest themselves in different temporal order. Thus, fusing the feature sets is challenging. We present a composite approach to affective sensing. The term composite is used to reflect the blending of information from multiple modalities with the available semantic evidence to enhance the emotion recognition process.

http://users.rsise.anu.edu.au/%7Eroland/Publications/McIntyreGoecke_LNCS4868_2008.pdf

Roland Göcke

Papers

Learning AAM fitting through simulation

An approach for automatically measuring facial activity in depressed subjects

Towards affective sensing

EREC-II in use: studies on usability and suitability of a sensor system for affect detection and human performance monitoring

The Composite Sensing of Affect