Head Nodding Research Articles

Toward an Agent Framework that Assists the Individuals with Mild Memory Impairment

With the increasing average life expectancy of world population, there are more and more dementia patients and the need of assistive technology emerges. According to the literature, the progress of cognitive impairment can be suppressed to be slower if the patients are constantly in calm mood. An effective way is suggested by keeping them in social relationship with the others. This paper presents a project with the goal of developing a framework for supporting dementia patients. This project features a conversational agent that can serve as a companion for dementia patients. Currently it has limited conversational ability but can generate back channel feedbacks, such as head nods and verbal acknowledgment, on the basis of acoustic information in the user's speech. In order to compensate the companionship of the agent and the ability to assist the user's memory, we are developing a memory vest which is equipped with portable devices including an Android smartphone, two IC audio recorders, and a digital video recorder to log the daily life of the patient. The gathered activity history database can then be used to enrich the dialogue ability of the agent and for helping the user to recall his / her own memory.

Shaken baby syndrome manifesting as infantile spasms seizure type

The diagnosis of child maltreatment leading to head injury is challenging. Here, we present the case of a 3-month-old female infant who presented with focal seizures that lasted for several minutes. After admission, she began to show intermittent clusters of head nods, irritable crying, arching, writhing, stiffening, and jerking of both arms. These results and electroencephalography findings were attributed as the diagnosis of infantile spasms (IS). Brain computed tomography and magnetic resonance imaging (MRI) revealed the presence of chronic subdural hematoma mixed with acute ischemic injuries. Examination of the eye fundus confirmed the presence of retinal hemorrhage. Therefore, all evidence pointed to a diagnosis of shaken baby syndrome (SBS). Based on this case, we suggest that physicians should consider a diagnosis of SBS for children with new-onset IS and that should be evaluated, diagnosed, and treated as promptly as possible.

Young children's understanding of denial.

Although a fair amount is known about young children's production of negation, little is known about their comprehension. Here, we focus on arguably the most complex basic form, denial, and how young children understand denial, when it is expressed in response to a question with gesture, single word, or sentence. One hundred twenty-six children in 3 age groups (Ms = 1 year 9 months, 2 years 0 months, and 2 years 4 months) witnessed an adult look into 1 of 2 buckets and then, in response to a question about whether the toy was in there, communicate either something positive (positive head nod, "yes," "it is in this bucket") or negative (negative head shake, "No," "It's not in this bucket"). The youngest children did not search differently in response to any of the communicative cues (nor in response to an additional cue using both gesture and single word). Children at 2 years 0 months searched at above-chance levels only in response to the negative word and negative sentence. Children at 2 years 4 months were successful with all 3 types of cues in both positive and negative modalities, with the exception of the positive sentence. Young children thus seem to understand the denial of a statement before they understand its affirmation, and they understand linguistic means of expressing denial before they understand gestural means.

Interactions between a quiz robot and multiple participants

This paper reports on a quiz robot experiment in which we explore similarities and differences in human participant speech, gaze, and bodily conduct in responding to a robot’s speech, gaze, and bodily conduct across two languages. Our experiment involved three-person groups of Japanese and English-speaking participants who stood facing the robot and a projection screen that displayed pictures related to the robot’s questions. The robot was programmed so that its speech was coordinated with its gaze, body position, and gestures in relation to transition relevance places (TRPs), key words, and deictic words and expressions (e.g. this, this picture) in both languages. Contrary to findings on human interaction, we found that the frequency of English speakers’ head nodding was higher than that of Japanese speakers in human-robot interaction (HRI). Our findings suggest that the coordination of the robot’s verbal and non-verbal actions surrounding TRPs, key words, and deictic words and expressions is important for facilitating HRI irrespective of participants’ native language. Keywords: coordination of verbal and non-verbal actions; robot gaze comparison between English and Japanese; human-robot interaction (HRI); transition relevance place (TRP); conversation analysis

Deterring predators, daunting opponents or drawing partners? Signaling rates across diverse contexts in the lizard Anolis sagrei

Although the visual display behavior in Anolis lizards has received ample attention, the function of dewlap extensions (DE), push-ups (PU), and head-nods (HN) in general, and in Anolis sagrei in particular, remains highly equivocal. Therefore, our primary goal was to quantify the display rates of these visual signal types (DE, PU, and HN) in a variety of contexts, using A. sagrei as study species. To our knowledge, this is the first study to test individuals of both sexes in a repeated-measures design across multiple contexts, including predator, non-predator and social interactions (mirror, male–male, male–female, female–male). We found that males have an overall higher signaling rate than females across all contexts. In addition, we found that lizards of both sexes exhibited higher display rates in the presence of conspecifics than when confronted with a predator or non-predator, suggesting that DE, PU, and HN function in intraspecific communication, not in predator deterrence. Whereas females did not significantly raise display rates in a consexual and heterosexual context with respect to subject-alone context, males did. The PU signal type only appears to play a major role for A. sagrei males during aggressive encounters. During heterosexual interactions, increased frequencies of all signal types suggest that DE, PU, and HN are essential for male courtship. Finally, we suggest that intersexual selection is probably a driving force for frequency-related dewlap use in both sexes. In contrast, pronounced intersexual differences were detected for PU and HN rates within a social context.

A longitudinal study on nodding syndrome—a new African epilepsy disorder

Nodding syndrome (NS), a new epilepsy disorder of sub-Saharan Africa, has only recently been classified. In a study conducted in southern Tanzania in 2005, 62 patients with NS were analyzed in great detail. The present study, a follow-up investigation, was conducted to evaluate the progression of NS over time and to obtain serial electroencephalography (EEG) data. Of the 62 NS patients, 53 (85.5%), the majority of whom were currently on some form of antiepileptic treatment, could be reevaluated in 2009 with a standardized questionnaire. A subset of these patients (25/53) underwent EEG investigation. In patients with "head nodding (HN) only" in 2005, 10 (43.5%) of 23 remained with the same diagnosis, whereas 5 (21.7%) of 23 had developed "HN plus" (i.e., HN and generalized tonic-clonic seizures). Six patients (26.1%) had seizures other than HN only, and two patients (8.7%) had fully recovered. In the "HN plus" group of 2005, 9 (30.0%) of 30 patients remained "HN plus," and 15 patients (50.0%) had seizures other than HN only. Four patients (13.3%) reverted to "HN only," and two patients (6.7%) stopped all seizures. In 11 (44.0%) of 25 patients, electroencephalography (EEG) showed generalized slowing. Six (54.6%) of these 11 abnormal EEG studies further showed generalized epileptiform discharges: (1) ictal electroencephalographic pattern with generalized 2.5 Hz spike and waves in two patients and (2) interictal bursts of 1.5-2 Hz spike and waves in four patients. This follow-up study confirms that HN represents an epilepsy disorder, possibly of the atypical absence type with dynamic development over time.

Benign and Periodic Movement Disorders in 2 Children With Down Syndrome

Children with Down syndrome show hypotonia and ligamentous laxity that are associated with motor development delay. Neurologic disorders are common in children with Down syndrome; however, in literature the presence of periodic movement disorders has not yet been described. We report 2 different types of periodic movement disorders in 2 infants with Down syndrome. In the first case, we described an 8-month-old girl with involuntary head nodding and absence of any other neurologic or ophthalmologic abnormalities. In the second case, we described a 6-month-old boy with abnormal but painless head rotation and inclination, alternating from side to side. Episodes of head tilting were often associated with a state of general uneasiness. Neurologic examination between attacks was normal. The aim of this paper is to provide practical information on recognition and management of movement disorders in Down syndrome.

825

Introduction: Communication impairment, as a result of mechanical ventilation, is a significant emotional stressor for critically ill patients and is a significant risk factor for preventable adverse events. The study purposes were to 1) describe interaction behaviors and factors (augmentative and Alternative Communication (AAC) strategies) that may influence communication between older patients and nurses and 2) explore associations between interaction behaviors and nursing care quality indicators (restraint use, pain management, sedation use and level). Methods: This expanded secondary study employed a descriptive correlational design utilizing data collected on a subset of 38 intubated, ICU patients > 60 years of age and their nurses (N=24) who participated in the Study of Patient-Nurse Effectiveness with Communication Strategies (SPEACS). Patients were: (1) intubated for at least 48 hours, (2) able to understand English, and (3) scored 13 or above on the Glasgow Coma Scale to be eligible for the SPEACS study. Behaviors were measured by rating videotaped observations between nurses and patients (four observations per dyad, total=152 observation) using the Communication Interaction Behavior Instrument (CIBI), which delineates positive and negative interaction behaviors of patients and nurses. Demographic and clinical characteristics and quality indicators were obtained from the SPEACS dataset and medical chart abstraction Descriptive statistics, group comparative statistics, and repeated measures analysis were utilized to explore the association between interaction behaviors and nursing care quality. Results: All positive behaviors occurred at least once; while negative behaviors were rare. The use of positive nurse behaviors was associated with an increase in positive patient behaviors (p<.001). The most frequently utilized strategies included: head nods, non-verbal but communicative action, gestures, and mouthing, The use of patient unaided AAC strategies was positively associated with the count of positive nurse behaviors (p=.002) and individual patient unaided AAC strategies were associated with individual nurse positive behaviors (p<0.05). Finally, count of different positive nurse behaviors was positively associated with the absence of reported pain (p=.011) and the count of different positive patient behaviors the association with the patient being calm (p<.001). Conclusions: Findings provide evidence that nurse and patient interaction behaviors impact the number and type of communication methods used by nonvocal patients to convey messages and may be associated with nursing care quality.

Activation of GABAergic Neurons in the Interpeduncular Nucleus Triggers Physical Nicotine Withdrawal Symptoms

Chronic exposure to nicotine elicits physical dependence in smokers, yet the mechanism and neuroanatomical bases for withdrawal symptoms are unclear. As in humans, rodents undergo physical withdrawal symptoms after cessation from chronic nicotine characterized by increased scratching, head nods, and body shakes. Here we show that induction of physical nicotine withdrawal symptoms activates GABAergic neurons within the interpeduncular nucleus (IPN). Optical activation of IPN GABAergic neurons via light stimulation of channelrhodopsin elicited physical withdrawal symptoms in both nicotine-naive and chronic-nicotine-exposed mice. Dampening excitability of GABAergic neurons during nicotine withdrawal through IPN-selective infusion of an NMDA receptor antagonist or through blockade of IPN neurotransmission from the medial habenula reduced IPN neuronal activation and alleviated withdrawal symptoms. During chronic nicotine exposure, nicotinic acetylcholine receptors containing the β4 subunit were upregulated in somatostatin interneurons clustered in the dorsal region of the IPN. Blockade of these receptors induced withdrawal signs more dramatically in nicotine-dependent compared to nicotine-naive mice and activated nonsomatostatin neurons in the IPN. Together, our data indicate that therapeutic strategies to reduce IPN GABAergic neuron excitability during nicotine withdrawal, for example, by activating nicotinic receptors on somatostatin interneurons, may be beneficial for alleviating withdrawal symptoms and facilitating smoking cessation.

Proposed guidelines for the management of nodding syndrome

Nodding Syndrome is a poorly understood neurologic disorder of unknown aetiology that affects children and adolescents in Africa. Recent studies have suggested that the head nods are due to atonic seizures and Nodding Syndrome may be classified as probably symptomatic generalised epilepsy. As part of the Ugandan Ministry of Health clinical management response, a multidisciplinary team developed a manual to guide the training of health workers with knowledge and skills to manage the patients. In the absence of a known cause, it was decided to offer symptomatic care. The objective is to relieve symptoms, offer primary and secondary prevention for disability and rehabilitation to improve function. Initial management focuses on the most urgent needs of the patient and the immediate family until 'stability' is achieved. The most important needs were considered as seizure control, management of behavioural and psychiatric difficulties, nursing care, nutritional and subsequently, physical and cognitive rehabilitation. This paper summarises the processes by which the proposed guidelines were developed and provides an outline of the specific treatments currently being provided for the patients.

MRI findings in people with epilepsy and nodding syndrome in an area endemic for onchocerciasis: An observational study

Onchocerciasis has been implicated in the pathogenesis of epilepsy. The debate on a potential causal relationship between Onchocerca volvulus and epilepsy has taken a new direction in the light of the most recent epidemic of nodding syndrome. To document MRI changes in people with different types of epilepsy and investigate whether there is an association with O. volvulus infection. In a prospective study in southern Tanzania, an area endemic for O. volvulus with a high prevalence of epilepsy and nodding syndrome, we performed MRI on 32 people with epilepsy, 12 of which suffered from nodding syndrome. Polymerase chain reaction (PCR) of O. volvulus was performed in skin and CSF. The most frequent abnormalities seen on MRI was atrophy (twelve patients (37.5%)) followed by intraparenchymal pathologies such as changes in the hippocampus (nine patients (28.1%)), gliotic lesions (six patients (18.8%)) and subcortical signal abnormalities (three patients (9.4%)). There was an overall trend towards an association of intraparenchymal cerebral pathologies and infection with O. volvulus based on skin PCR (Fisher's Exact Test p=0.067) which was most pronounced in children and adolescents with nodding syndrome compared to those with other types of epilepsy (Fisher's Exact Test, p=0.083). Contrary to skin PCR results, PCR of CSF was negative in all patients. The observed trend towards an association of intraparenchymal cerebral pathological results on MRI and a positive skin PCR for O. volvulus despite negative PCR of CSF is intriguing and deserves further attention.

Nodding Syndrome

An epidemic illness characterized by head nodding associated with onchocerciasis has been described in eastern Africa since the early 1960s; we summarize published reports and recent studies. Onset of nodding occurs in previously healthy 5–15-year-old children and is often triggered by eating or cold temperatures and accompanied by cognitive impairment. Its incidence has increased in Uganda and South Sudan over the past 10 years. Four case–control studies identified modest and inconsistent associations. There were nonspecific lesions seen by magnetic resonance imaging, no cerebrospinal fluid inflammation, and markedly abnormal electroencephalography results. Nodding episodes are atonic seizures. Testing has failed to demonstrate associations with trypanosomiasis, cysticercosis, loiasis, lymphatic filariasis, cerebral malaria, measles, prion disease, or novel pathogens; or deficiencies of folate, cobalamin, pyridoxine, retinol, or zinc; or toxicity from mercury, copper, or homocysteine. There is a consistent enigmatic association with onchocerciasis detected by skin snip or serologic analysis. Nodding syndrome is an unexplained epidemic epilepsy.

Visual Analysis of Eye State and Head Pose for Driver Alertness Monitoring

This paper presents visual analysis of eye state and head pose (HP) for continuous monitoring of alertness of a vehicle driver. Most existing approaches to visual detection of nonalert driving patterns rely either on eye closure or head nodding angles to determine the driver drowsiness or distraction level. The proposed scheme uses visual features such as eye index (EI), pupil activity (PA), and HP to extract critical information on nonalertness of a vehicle driver. EI determines if the eye is open, half closed, or closed from the ratio of pupil height and eye height. PA measures the rate of deviation of the pupil center from the eye center over a time period. HP finds the amount of the driver's head movements by counting the number of video segments that involve a large deviation of three Euler angles of HP, i.e., nodding, shaking, and tilting, from its normal driving position. HP provides useful information on the lack of attention, particularly when the driver's eyes are not visible due to occlusion caused by large head movements. A support vector machine (SVM) classifies a sequence of video segments into alert or nonalert driving events. Experimental results show that the proposed scheme offers high classification accuracy with acceptably low errors and false alarms for people of various ethnicity and gender in real road driving conditions.

P 23. Safety aspects of anodal transcranial direct current stimulation: Dose–response effects on EEG and sensory evoked potentials

Introduction Anodal transcranial direct-current stimulation (tDCS) has been shown to modulate cortical excitability and to improve motor learning (Nitsche and Paulus, 2000; Reis et al., 2009). The human studies on tDCS generally use low stimulation intensities-approximately 0.4–0.8A/m 2 -to ensure safe levels. However, the dose–response relationship of tDCS on functional measures of cortical excitability and possible adverse effects of tDCS have not yet been investigated. Objective In this in vivo animal model of tDCS dose-dependent effects of anodal tDCS on cortical excitability were investigated and safety limits for safe application were determined. Methods Adult male Sprague–Dawley rats were equipped with electrodes for tDCS stimulation at the left primary motor cortex and the chest region. Furthermore, EEG electrodes were surgically placed above the somatosensory cortices and the cerebellum (grounding electrode). In separate sessions, the rats were then exposed to anodal tDCS of different intensities (between 2 and 31.8A/m 2 ) for 20min. EEG and sensory evoked potentials were continuously recorded prior to, during and after tDCS, in freely moving (EEG) or anaesthetised (SEP) animals. Results EEG data clearly showed that low stimulation intensities did not affect EEG patterns. Mild EEG alterations-showing more spiky waves and increased amplitudes-were present at the lowest stimulation intensity of 2A/m 2 , while at an intensity of 31.8A/m 2 EEG seizure patterns were observed, also associated with clinical seizures (facial twitch, head nodding, i.e. stage 1–2 on the Racine seizure scale). The calculated threshold (ED1, effective dose 1%) for first occurrence of EEG alterations was 1.35A/m 2 , the ED50 was estimated at 9.2A/m 2 . SEP data are currently under investigation. Conclusions Since only a high stimulation intensity, approximately 40 times higher than that applied to humans, was found to cause seizures, the stimulation intensities applied in human studies are unlikely to be harmful. This suggests that higher stimulation intensities could be applied in human studies, thereby possibly further improving plasticity, i.e. learning, without putting the brain at risk.

A Nod in the Wrong Direction: Does Non-verbal Feedback Affect Eyewitness Confidence in Interviews?

Eyewitnesses can be influenced by an interviewer's behaviour and report information with inflated confidence as a result. Previous research has shown that positive feedback administered verbally can affect the confidence attributed to testimony, but the effect of non-verbal influence in interviews has been given little attention. This study investigated whether positive or negative non-verbal feedback could affect the confidence witnesses attribute to their responses. Participants witnessed staged CCTV footage of a crime scene and answered 20 questions in a structured interview, during which they were given either positive feedback (a head nod), negative feedback (a head shake) or no feedback. Those presented with positive non-verbal feedback reported inflated confidence compared with those presented with negative non-verbal feedback regardless of accuracy, and this effect was most apparent when participants reported awareness of the feedback. These results provide further insight into the effects of interviewer behaviour in investigative interviews.

Determinants of Oxygen Therapy in Childhood Pneumonia in a Resource-Constrained Region

Childhood pneumonia is a leading cause of morbidity and mortality among underfives particularly in the resource-constraint part of the world. A high proportion of these deaths are due to lack of oxygen, thereby making oxygen administration a life-saving adjunctive when indicated. However, many primary health centres that manage most of the cases often lack the adequate manpower and facilities to decide which patient should be on oxygen therapy. Therefore, this study aimed to determine factors that predict hypoxaemia at presentation in children with severe pneumonia. Four hundred and twenty children aged from 2 to 59 months (40% infants) with severe pneumonia admitted to a health centre in rural Gambia were assessed at presentation. Eighty-one of them (19.30%) had hypoxaemia (oxygen saturation < 90%). Children aged 2–11 months, with grunting respiration, cyanosis, and head nodding, and those with cardiomegaly on chest radiograph were at higher risk of hypoxaemia (P < 0.05). Grunting respiration (OR = 5.210, 95% CI 2.287–7.482) and cyanosis (OR = 83.200, 95% CI 5.248–355.111) were independent predictors of hypoxaemia in childhood pneumonia. We conclude that children that grunt and are centrally cyanosed should be preferentially commenced on oxygen therapy even when there is no facility to confirm hypoxaemia.

Development of head detection and tracking systems for visual surveillance

This paper proposes a technique for the detection of head nod and shake gestures based on eye tracking and head motion decision. The eye tracking step is divided into face detection and eye location. Here, we apply a motion segmentation algorithm that examines differences in moving people's faces. This system utilizes a Hidden Markov Model-based head detection module that carries out complete detection in the input images, followed by the eye tracking module that refines the search based on a candidate list provided by the preprocessing module. The novelty of this paper is derived from differences in real-time input images, preprocessing to remove noises (morphological operators and so on), detecting edge lines and restoration, finding the face area, and cutting the head candidate. Moreover, we adopt a K-means algorithm for finding the head region. Real-time eye tracking extracts the location of eyes from the detected face region and is performed at close to a pair of eyes. After eye tracking, the coordinates of the detected eyes are transformed into a normalized vector of x-coordinate and y-coordinate. Head nod and shake detector uses three hidden Markov models (HMMs). HMM representation of the head detection can estimate the underlying HMM states from a sequence of face images. Head nod and shake can be detected by three HMMs that are adapted by a directional vector. The directional vector represents the direction of the head movement. The vector is HMMs for determining neutral as well as head nod and shake. These techniques are implemented on images, and notable success is notified.

The Auxiliary Subunit KChIP2 Is an Essential Regulator of Homeostatic Excitability

The necessity for, or redundancy of, distinctive KChIP proteins is not known. Deletion of KChIP2 leads to increased susceptibility to epilepsy and to a reduction in IA and increased excitability in pyramidal hippocampal neurons. KChIP2 is essential for homeostasis in hippocampal neurons. Mutations in K(A) channel auxiliary subunits may be loci for epilepsy. The somatodendritic IA (A-type) K(+) current underlies neuronal excitability, and loss of IA has been associated with the development of epilepsy. Whether any one of the four auxiliary potassium channel interacting proteins (KChIPs), KChIP1-KChIP4, in specific neuronal populations is critical for IA is not known. Here we show that KChIP2, which is abundantly expressed in hippocampal pyramidal cells, is essential for IA regulation in hippocampal neurons and that deletion of Kchip2 affects susceptibility to limbic seizures. The specific effects of Kchip2 deletion on IA recorded from isolated hippocampal pyramidal neurons were a reduction in amplitude and shift in the V½ for steady-state inactivation to hyperpolarized potentials when compared with WT neurons. Consistent with the relative loss of IA, hippocampal neurons from Kchip2(-/-) mice showed increased excitability. WT cultured neurons fired only occasional single action potentials, but the average spontaneous firing rate (spikes/s) was almost 10-fold greater in Kchip2(-/-) neurons. In slice preparations, spontaneous firing was detected in CA1 pyramidal neurons from Kchip2(-/-) mice but not from WT. Additionally, when seizures were induced by kindling, the number of stimulations required to evoke an initial class 4 or 5 seizure was decreased, and the average duration of electrographic seizures was longer in Kchip2(-/-) mice compared with WT controls. Together, these data demonstrate that the KChIP2 is essential for physiologic IA modulation and homeostatic stability and that there is a lack of functional redundancy among the different KChIPs in hippocampal neurons.

Assessing the Validity of the Dynamic Gait Index in a Balance Disorders Clinic: An Application of Rasch Analysis

The Dynamic Gait Index (DGI) has emerged as a valid indicator of functional gait abilities for people with balance and vestibular disorders. Recent Rasch-based analyses have indicated possible concerns for multidimensionality and a ceiling effect within the DGI. The aim of this study was to evaluate the DGI in a sample of patients from a dizziness and balance clinic to determine whether patient features such as dizziness or fall history influence the measurement characteristics of the DGI. This study used a retrospective design. A sample of 117 patients' charts was reviewed, and patients were grouped according to a primary impairment of dizziness only or imbalance and were categorized based on a history of falls. A one-parameter Rasch-Andrich rating scale model was used with thorough analyses, including rating scale analysis, item-difficulty hierarchy, scale unidimensionality, and differential item functioning (DIF). The DGI demonstrated an effective rating scale design and was found to be a unidimensional measurement of dynamic gait. The DGI displayed a modest ceiling effect, primarily with patients with higher functional levels displaying symptoms of dizziness. Three items ("vertical head nods," "gait on level surface," and "stepping over obstacles") demonstrated DIF based on categories of patient characteristics, although the effects on measurement were negligible. Functional categories were based on impairments and not underlying medical diagnoses derived from a retrospective chart review, whereas the limited sample size may have underestimated statistically significant DIF. Results from this study offer additional evidence supporting the validity of the DGI as a measure of gait ability. The present findings also are in agreement with prior research that has shown a ceiling effect for the DGI in people with balance or vestibular disorders. Effects of DIF were found to be negligible, yet the presence of DIF within the present sample helped to explain some differences in DGI item-difficulty hierarchies from prior studies. Continued research is needed to determine how population differences may affect performance on the DGI and to develop and test assessments capable of measuring a broader range of gait abilities.

Latent Mixture of Discriminative Experts

In this paper, we introduce a new model called Latent Mixture of Discriminative Experts which can automatically learn the temporal relationship between different modalities. Since, we train separate experts for each modality, LMDE is capable of improving the prediction performance even with limited amount of data. For model interpretation, we present a sparse feature ranking algorithm that exploits <i xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">L</i> <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sub> regularization. An empirical evaluation is provided on the task of listener backchannel prediction (i.e., head nod). We introduce a new error evaluation metric called User-adaptive Prediction Accuracy that takes into account the difference in people's backchannel responses. Our results confirm the importance of combining five types of multimodal features: lexical, syntactic structure, part-of-speech, visual and prosody. Latent Mixture of Discriminative Experts model outperforms previous approaches.