Head Orientation Research Articles

Multimodal Joint Head Orientation Estimation in Interacting Groups via Proxemics and Interaction Dynamics

Human head orientation estimation has been of interest because head orientation serves as a cue to directed social attention. Most existing approaches rely on visual and high-fidelity sensor inputs and deep learning strategies that do not consider the social context of unstructured and crowded mingling scenarios. We show that alternative inputs, like speaking status, body location, orientation, and acceleration contribute towards head orientation estimation. These are especially useful in crowded and in-the-wild settings where visual features are either uninformative due to occlusions or prohibitive to acquire due to physical space limitations and concerns of ecological validity. We argue that head orientation estimation in such social settings needs to account for the physically evolving interaction space formed by all the individuals in the group. To this end, we propose an LSTM-based head orientation estimation method that combines the hidden representations of the group members. Our framework jointly predicts head orientations of all group members and is applicable to groups of different sizes. We explain the contribution of different modalities to model performance in head orientation estimation. The proposed model outperforms baseline methods that do not explicitly consider the group context, and generalizes to an unseen dataset from a different social event.

Effect of Le Fort I Maxillary Repositioning on Three-Dimensional Nasal Tip Rotation: A Comparative Study with Implication for the Asian Nose

Background: Le Fort I maxillary repositioning influences nasal morphology. In Asian cultures, upward nasal tip rotation with increased nostril exposure is considered aesthetically unpleasant and can have psychosocial consequences. This three-dimensional imaging–based study evaluated the effect of different Le Fort I maxillary movements on nasal tip rotation. Methods: Consecutive patients who underwent two-jaw orthognathic surgery (n = 107) were enrolled. To achieve a standard head orientation, preoperative and 1-week and 12-month postoperative cone-beam computed tomography–derived three-dimensional craniofacial models were superimposed. Tip rotation angle was calculated according to the Frankfort horizontal plane for all three-dimensional digital models. The final tip rotation angle change was defined as 12-month postoperative value minus preoperative value. Translational maxillary movement types (advancement versus setback and intrusion versus extrusion), postoperative maxillary segment locations (anterosuperior, anteroinferior, posterosuperior, or posteroinferior), and actual linear maxillary changes were noted. Results: Advancement (1.79 ± 5.20 degrees) and intrusion (2.23 ± 4.96 degrees) movements demonstrated significantly larger final tip rotation angle changes than setback (−0.88 ± 5.15 degrees) and extrusion (0.09 ± 5.44 degrees) movements (all p < 0.05). Postoperative anterosuperior location (2.95 ± 4.52 degrees) of the maxillary segment demonstrated a significantly larger final tip rotation angle change than anteroinferior (0.48 ± 5.65 degrees), posterosuperior (−1.08 ± 4.77 degrees), and posteroinferior (−0.64 ± 5.80 degrees) locations (all p < 0.05). Translational maxillary movement and actual linear maxillary change were not correlated with final tip rotation angle change. Conclusion: Effects of Le Fort I maxillary repositioning on nasal tip rotation depend on movement types and maxillary segment location. CLINICAL QUESTION/LEVEL OF EVIDENCE: Therapeutic, III.

An Implanted Vestibular Prosthesis Improves Spatial Orientation in Animals with Severe Vestibular Damage

Gravity is a pervasive environmental stimulus, and accurate graviception is required for optimal spatial orientation and postural stability. The primary graviceptors are the vestibular organs, which include angular velocity (semicircular canals) and linear acceleration (otolith organs) sensors. Graviception is degraded in patients with vestibular damage, resulting in spatial misperception and imbalance. Since minimal therapy is available for these patients, substantial effort has focused on developing a vestibular prosthesis or vestibular implant (VI) that reproduces information normally provided by the canals (since reproducing otolith function is very challenging technically). Prior studies demonstrated that angular eye velocity responses could be driven by canal VI-mediated angular head velocity information, but it remains unknown whether a canal VI could improve spatial perception and posture since these behaviors require accurate estimates of angular head position in space relative to gravity. Here, we tested the hypothesis that a canal VI that transduces angular head velocity and provides this information to the brain via motion-modulated electrical stimulation of canal afferent nerves could improve the perception of angular head position relative to gravity in monkeys with severe vestibular damage. Using a subjective visual vertical task, we found that normal female monkeys accurately sensed the orientation of the head relative to gravity during dynamic tilts, that this ability was degraded following bilateral vestibular damage, and improved when the canal VI was used. These results demonstrate that a canal VI can improve graviception in vestibulopathic animals, suggesting that it could reduce the disabling perceptual and postural deficits experienced by patients with severe vestibular damage.SIGNIFICANCE STATEMENTPatients with vestibular damage experience impaired vision, spatial perception, and balance, symptoms that could potentially respond to a vestibular implant (VI). Anatomic features facilitate semicircular canal (angular velocity) prosthetics but inhibit approaches with the otolith (linear acceleration) organs, and canal VIs that sense angular head velocity can generate compensatory eye velocity responses in vestibulopathic subjects. Can the brain use canal VI head velocity information to improve estimates of head orientation (e.g., head position relative to gravity), which is a prerequisite for accurate spatial perception and posture? Here we show that a canal VI can improve the perception of head orientation in vestibulopathic monkeys, results that are highly significant because they suggest that VIs mimicking canal function can improve spatial orientation and balance in vestibulopathic patients.

Using a new video rating tool to crowd-source analysis of behavioural reaction to stimuli

Quantifying the intensity of animals’ reaction to stimuli is notoriously difficult as classic unidimensional measures of responses such as latency or duration of looking can fail to capture the overall strength of behavioural responses. More holistic rating can be useful but have the inherent risks of subjective bias and lack of repeatability. Here, we explored whether crowdsourcing could be used to efficiently and reliably overcome these potential flaws. A total of 396 participants watched online videos of dogs reacting to auditory stimuli and provided 23,248 ratings of the strength of the dogs’ responses from zero (default) to 100 using an online survey form. We found that raters achieved very high inter-rater reliability across multiple datasets (although their responses were affected by their sex, age, and attitude towards animals) and that as few as 10 raters could be used to achieve a reliable result. A linear mixed model applied to PCA components of behaviours discovered that the dogs’ facial expressions and head orientation influenced the strength of behaviour ratings the most. Further linear mixed models showed that that strength of behaviour ratings was moderately correlated to the duration of dogs’ reactions but not to dogs’ reaction latency (from the stimulus onset). This suggests that observers’ ratings captured consistent dimensions of animals’ responses that are not fully represented by more classic unidimensional metrics. Finally, we report that overall participants strongly enjoyed the experience. Thus, we suggest that using crowdsourcing can offer a useful, repeatable tool to assess behavioural intensity in experimental or observational studies where unidimensional coding may miss nuance, or where coding multiple dimensions may be too time-consuming.

Verification of fetal evoked response by magnetic dipole fitting

BackgroundInvestigation of fetal evoked response to auditory or visual stimuli is an important means of understanding the developmental stages and potential problems in prenatal life. It is, however, not without certain imperfections. The biggest challenge with fetal evoked response is its low signal to noise ratio. Under noisy conditions, the detected fetal evoked response should, therefore, be further investigated to confirm that the source of the signal is from fetal brain and is not related to random noise. Existing methods for verification are: (1) visual inspection of magnetic field maps, which requires user intervention and expert knowledge which can be highly subjective; (2) simultaneous ultrasound measurement, which is expensive and technically difficult to manage; and (3) equivalent current dipole fitting, which requires knowledge of the orientation of fetal head and its dimensions that may not be available at all times. ObjectiveTo verify that the detected fetal evoked response signal is originating from the fetal head by using an objective and feasible method that employs magnetic dipole fitting to fetal evoked response. Study designFrom raw fetal magnetoencephalography data, the cardiac interference was removed by frequency dependent subtraction. After averaging over stimulus triggers, the resulting signal was taken as the candidate fetal evoked response. The fetal evoked response was investigated for the highest peak in between 0.2–0.5 s, which is the expected latency of the response to the stimulus. The magnetic field at this highest peak was used for magnetic dipole fitting. The validation of peak was based on the closeness of the magnetic dipole fit to vicinity of fetal head location determined by ultrasound and the anatomically reasonable distance from the fetal heart. The methodology was first tested on a sample neonatal data before application to fetal data. ResultsThe results of neonatal application confirmed that the source localization by magnetic dipole fitting for the brain produced meaningful results. When applied to fetal data, auditory and visual evoked response was detected in 27 of the 38 recordings. This implied that with our verification method, fetal evoked responses were detected in 71% of fetuses. ConclusionDetection rate of the evoked responses were similar to earlier reports where subjective visual inspection or simultaneous ultrasound measurement were used. Our method using magnetic dipole fitting for verification is more feasible and objective compared to the earlier methods.

Forecasting People Trajectories and Head Poses by Jointly Reasoning on Tracklets and Vislets.

In this article, we explore the correlation between people trajectories and their head orientations. We argue that people trajectory and head pose forecasting can be modelled as a joint problem. Recent approaches on trajectory forecasting leverage short-term trajectories (aka tracklets) of pedestrians to predict their future paths. In addition, sociological cues, such as expected destination or pedestrian interaction, are often combined with tracklets. In this article, we propose MiXing-LSTM (MX-LSTM) to capture the interplay between positions and head orientations (vislets) thanks to a joint unconstrained optimization of full covariance matrices during the LSTM backpropagation. We additionally exploit the head orientations as a proxy for the visual attention, when modeling social interactions. MX-LSTM predicts future pedestrians location and head pose, increasing the standard capabilities of the current approaches on long-term trajectory forecasting. Compared to the state-of-the-art, our approach shows better performances on an extensive set of public benchmarks. MX-LSTM is particularly effective when people move slowly, i.e., the most challenging scenario for all other models. The proposed approach also allows for accurate predictions on a longer time horizon.

Laterality and region-specific tau phosphorylation correlate with PTSD-related behavioral traits in rats exposed to repetitive low-level blast

Military veterans who experience blast-related traumatic brain injuries often suffer from chronic cognitive and neurobehavioral syndromes. Reports of abnormal tau processing following blast injury have raised concerns that some cases may have a neurodegenerative basis. Rats exposed to repetitive low-level blast exhibit chronic neurobehavioral traits and accumulate tau phosphorylated at threonine 181 (Thr181). Using data previously reported in separate studies we tested the hypothesis that region-specific patterns of Thr181 phosphorylation correlate with behavioral measures also previously determined and reported in the same animals. Elevated p-tau Thr181 in anterior neocortical regions and right hippocampus correlated with anxiety as well as fear learning and novel object localization. There were no correlations with levels in amygdala or posterior neocortical regions. Particularly striking were asymmetrical effects on the right and left hippocampus. No systematic variation in head orientation toward the blast wave seems to explain the laterality. Levels did not correlate with behavioral measures of hyperarousal. Results were specific to Thr181 in that no correlations were observed for three other phospho-acceptor sites (threonine 231, serine 396, and serine 404). No consistent correlations were linked with total tau. These correlations are significant in suggesting that p-tau accumulation in anterior neocortical regions and the hippocampus may lead to disinhibited amygdala function without p-tau elevation in the amygdala itself. They also suggest an association linking blast injury with tauopathy, which has implications for understanding the relationship of chronic blast-related neurobehavioral syndromes in humans to neurodegenerative diseases.

Variation between natural head orientation and Frankfort horizontal planes in orthognathic surgery patients: 187 consecutive cases

The purpose of this study was to assess the relationship between the Frankfort horizontal (FH) and natural head orientation (NHO), their correlation between patients’ malocclusion, and the impact of counterclockwise rotation (CCW) on the FH-NHO angle variation after orthognathic surgery. An evaluation of 187 consecutive patients was performed at the Maxillofacial Institute (Teknon Medical Center, Barcelona). FH-NHO° was measured pre- and postoperatively at 1 and 12 months, after three-dimensional (3D) superimposition using a software (Dolphin®). Patients were classified as follows: 3.2%, 48.7% and 48.1%, class I, II and III, respectively. Baseline FH-NHO° was significantly positive for patients with dentofacial deformities (2.73°±4.19 (2.12–3.33°, P<0.001). The impact of orthognathic surgery in FH-NHO° was greater in class II when compared with class III patients, with a variation of 2.04°±4.79 (P<0.001) and −1.20°±3.03 (P<0.001), respectively. FH-NHO° increased when CCW rotational movements were performed (P=0.006). The results of this study suggest that pre- and postoperative NHO differs from FH in orthognathic patients. The angle between FH and NHO is significantly larger in class III than in class II patients at baseline, which converges after orthognathic surgery when CCW rotation is performed. Therefore, NHO should be used as the real horizontal plane when planning for orthognathic surgery.

Auditory perception is influenced by the orientation of the trunk relative to a sound source.

The study investigated how hearing depends on the whole body, head and trunk orientation relative to a sound source. In normal hearing humans we examined auditory thresholds and their ability to recognize logatomes (bi-syllabic non-sense words) at different whole body, head and trunk rotation relative to a sound source. We found that auditory threshold was increased and logatome recognition was impaired when the body or the trunk were rotated 40° away from a sound source compared to when the body or the trunk was oriented towards the sound source. Conversely, no effects were seen when only the head was rotated. Further, an increase of thresholds and impairment of logatome recognition were also observed after unilateral vibration of dorsal neck muscles that induces, per se, long-lasting illusory trunk displacement relative to the head. Thus, our findings support the idea that processing of acoustic signals depends on where a sound is located within a reference system defined by the subject's trunk coordinates.

Listener-Position and Orientation Dependency of Auditory Perception in an Enclosed Space: Elicitation of Salient Attributes

This paper presents a subjective study conducted on the perception of auditory attributes depending on listener position and head orientation in an enclosed space. Two elicitation experiments were carried out using the repertory grid technique—in-situ and laboratory experiments—which aimed to identify perceptual attributes among 10 different combinations of the listener’s positions and head orientations in a concert hall. It was found that, between the in-situ and laboratory experiments, the listening positions and head orientations were clustered identically. Ten salient perceptual attributes were identified from the data obtained from the laboratory experiment. Whilst these included conventional attributes such as ASW (apparent source width) and LEV (listener envelopment), new attributes such as PRL (perceived reverb loudness), ARW (apparent reverb width) and Reverb Direction were identified, and they are hypothesised to be sub-attributes of LEV (listener envelopment). Timbral characteristics such as Reverb Brightness and Echo Brightness were also identified as salient attributes, which are considered to potentially contribute to the overall perceived clarity.

Effect of camera position on egocentric localisation with Simulated Prosthetic Vision

Objective: The positioning of the external camera in current designs of retinal prostheses may affect the egocentric spatial frame of reference and lead to pointing and localisation errors. The present simulated prosthetic vision (SPV) study investigated the immediate effects of four camera positions: (i) centre-mounted (CM), (ii) head-mounted (HM), (iii) left-mounted (LM), (iv) right-mounted (RM) on egocentric localisation. Approach: Fifteen sighted participants performed a pointing task while viewing the stimulus through a simulation of Bionic Vision Technologies’ 44-channel retinal prosthesis. Pointing responses were recorded immediately after exposure to different camera positions and prior to adaptation. The direction and magnitude of pointing errors and associated head movement behaviour were analysed using linear mixed effects analysis. Main results: Our results showed lower errors for medial camera positions (CM and HM) and overall improved accuracy and precision for CM position. Head orientation was found to largely influence the direction of pointing. There was a high positive correlation between direction of head orientation and direction of pointing with CM and HM camera positions, suggesting medial camera positions facilitated intuitive head orientation cues that are important for reliable egocentric direction estimation. Significance: The empirical evidence derived from the present SPV study offers practical guidance to reduce the effect on spatial dissociation of camera position in localisation tasks and inform design of future prosthetic vision devices and improve rehabilitation approaches.

Simulation as a Training Method for Electricity Workers' Safety.

Background: Simulation is a useful method to improve learning and increase the safety of work operations, both for technical and non-technical skills. However, the observation, assessment, and feedback about these skills is particularly complex, because the process needs expert observers, and the feedback could be judgmental and ineffective. Therefore, a structured process to develop effective simulation scenarios and tools for the observation and feedback about performance is crucial. To this aim, in the present research, we developed a training model for electricity distribution workers, based on high fidelity simulation. Methods: We designed simulation scenarios based on real cases, developed, and tested a set of observation and rating forms for the non-technical skills behavioral markers, and we tracked behaviors based on non-verbal cues (physiological and head orientation parameters). Results: The training methodology proved to be highly appreciated by the participants and effective in fostering reflexivity. An in-depth analysis of physiological indexes and behaviors compliant to safety procedures revealed that breath rate and heart rate patterns commonly related with mindful and relaxed states were correlated with compliant behaviors, and patterns typical of stress and anxiety were correlated with non-compliant behaviors. Conclusions: a new training method based on high fidelity simulation, addressing both technical and non-technical skills is now available for fostering self-reflection and safety for electricity distribution workers. Future research should assess the long-term effectiveness of high-fidelity simulation for electricity workers, and should investigate non-invasive and real-time methods for tracking physiological parameters.

Head motion classification using thread-based sensor and machine learning algorithm

Human machine interfaces that can track head motion will result in advances in physical rehabilitation, improved augmented reality/virtual reality systems, and aid in the study of human behavior. This paper presents a head position monitoring and classification system using thin flexible strain sensing threads placed on the neck of an individual. A wireless circuit module consisting of impedance readout circuitry and a Bluetooth module records and transmits strain information to a computer. A data processing algorithm for motion recognition provides near real-time quantification of head position. Incoming data is filtered, normalized and divided into data segments. A set of features is extracted from each data segment and employed as input to nine classifiers including Support Vector Machine, Naive Bayes and KNN for position prediction. A testing accuracy of around 92% was achieved for a set of nine head orientations. Results indicate that this human machine interface platform is accurate, flexible, easy to use, and cost effective.

Estimation of Driver Lane Change Intention Based on the LSTM and Dempster–Shafer Evidence Theory

Rapid and correct estimation of driver lane change intention plays an important role in the advanced driver assistance system (ADAS), which could make the driver improve the reliability of the ADAS system and help to decrease driver workload. In this study, a method based on the long short-term memory network (LSTM) and Dempster–Shafer evidence theory is proposed. The model consists of a preliminary decision-making label and a final decision-making label. Driver visual information, head orientation, and vehicle dynamics are collected by preliminary decision-making label. Then, LSTM is used to calculate the initial probability of the driver lane change (left, right, and lane keeping) maneuver intention. The outputs of LSTM are normalized and assigned a basic probability by the Dempster–Shafer evidence theory. The final decision-making label analyzes the information and outputs the probability of each lane change intention and the decision is to identify the driver's current intention. The experimental results show that the accuracy of the model is 90.7% for the intention of changing left and 89.1% for the intention of changing right. The outcome of this work is an essential component for all levels of road vehicle automation.

The Critical Role of Head Movements for Spatial Representation During Bumblebees Learning Flight.

Bumblebees perform complex flight maneuvers around the barely visible entrance of their nest upon their first departures. During these flights bees learn visual information about the surroundings, possibly including its spatial layout. They rely on this information to return home. Depth information can be derived from the apparent motion of the scenery on the bees' retina. This motion is shaped by the animal's flight and orientation: Bees employ a saccadic flight and gaze strategy, where rapid turns of the head (saccades) alternate with flight segments of apparently constant gaze direction (intersaccades). When during intersaccades the gaze direction is kept relatively constant, the apparent motion contains information about the distance of the animal to environmental objects, and thus, in an egocentric reference frame. Alternatively, when the gaze direction rotates around a fixed point in space, the animal perceives the depth structure relative to this pivot point, i.e., in an allocentric reference frame. If the pivot point is at the nest-hole, the information is nest-centric. Here, we investigate in which reference frames bumblebees perceive depth information during their learning flights. By precisely tracking the head orientation, we found that half of the time, the head appears to pivot actively. However, only few of the corresponding pivot points are close to the nest entrance. Our results indicate that bumblebees perceive visual information in several reference frames when they learn about the surroundings of a behaviorally relevant location.

Analysing Head-Thorax Choreography During Free-Flights in Bumblebees.

Animals coordinate their various body parts, sometimes in elaborate manners to swim, walk, climb, fly, and navigate their environment. The coordination of body parts is essential to behaviors such as, chasing, escaping, landing, and the extraction of relevant information. For example, by shaping the movement of the head and body in an active and controlled manner, flying insects structure their flights to facilitate the acquisition of distance information. They condense their turns into a short period of time (the saccade) interspaced by a relatively long translation (the intersaccade). However, due to technological limitations, the precise coordination of the head and thorax during insects' free-flight remains unclear. Here, we propose methods to analyse the orientation of the head and thorax of bumblebees Bombus terrestris, to segregate the trajectories of flying insects into saccades and intersaccades by using supervised machine learning (ML) techniques, and finally to analyse the coordination between head and thorax by using artificial neural networks (ANN). The segregation of flights into saccades and intersaccades by ML, based on the thorax angular velocities, decreased the misclassification by 12% compared to classically used methods. Our results demonstrate how machine learning techniques can be used to improve the analyses of insect flight structures and to learn about the complexity of head-body coordination. We anticipate our assay to be a starting point for more sophisticated experiments and analysis on freely flying insects. For example, the coordination of head and body movements during collision avoidance, chasing behavior, or negotiation of gaps could be investigated by monitoring the head and thorax orientation of freely flying insects within and across behavioral tasks, and in different species.

Effects of Reinforcement Learning on Gaze Following of Gaze and Head Direction in Early Infancy: An Interactive Eye-Tracking Study.

The current four experiments investigated gaze following behavior in response to gaze and head turns in 4-month-olds and how reinforcement learning influences this behavior (N=99). Using interactive eye tracking, infants' gaze elicited an animation whenever infants followed a person's head or gaze orientation (Experiment 1.1, 2.1 and 2.2) or looked at the opposite side (Experiment 1.2). Infants spontaneously followed the direction of a turning head with and without simultaneously shifted gaze direction (Cohen's d: 0.93-1.05) but not the direction of isolated gaze shifts. We only found a weak effect of reinforcement on gaze following in one of the four experiments. Results will be discussed with regard to the impact of reinforcement on the maintenance of already existing gaze following behavior.

Evaluation of head-tracked binaural auralizations of speech signals generated with a virtual artificial head in anechoic and classroom environments

In order to realize binaural auralizations with head tracking, BRIRs of individual listeners are needed for different head orientations. In this contribution, a filter-and-sum beamformer, referred to as virtual artificial head (VAH), was used to synthesize the BRIRs. To this end, room impulse responses were first measured with a VAH, using a planar microphone array with 24 microphones, for one fixed orientation, in an anechoic and a reverberant room. Then, individual spectral weights for 185 orientations of the listener’s head were calculated with different parameter sets. Parameters included the number and the direction of the sources considered in the calculation of spectral weights as well as the required minimum mean white noise gain (WNGm). For both acoustical environments, the quality of the resulting synthesized BRIRs was assessed perceptually in head-tracked auralizations, in direct comparison to real loudspeaker playback in the room. Results showed that both rooms could be auralized with the VAH for speech signals in a perceptually convincing manner, by employing spectral weights calculated with 72 source directions from the horizontal plane. In addition, low resulting WNGmvalues should be avoided. Furthermore, in the dynamic binaural auralization with speech signals in this study, individual BRIRs seemed to offer no advantage over non-individual BRIRs, confirming previous results that were obtained with simulated BRIRs.

Contribution of Basal Ganglia to the Sense of Upright: A Double-Blind Within-Person Randomized Trial of Subthalamic Stimulation in Parkinson's Disease with Pisa Syndrome.

Verticality perception is frequently altered in Parkinson's disease (PD) with Pisa syndrome (PS). Is it the cause or the consequence of the PS? We tested the hypothesis that both scenarios coexist. We performed a double-blind within-person randomized trial (NCT02704910) in 18 individuals (median age 63.5 years) with PD evolving for a median of 17.5 years and PS for 2.5 years and treated with bilateral stimulation of the subthalamus nuclei (STN-DBS) for 6.5 years. We analyzed whether head and trunk orientations were congruent with the visual (VV) and postural (PV) vertical, and whether switching on one or both sides of the STN-DBS could modulate trunk orientation via verticality representation. The tilted verticality perception could explain the PS in 6/18 (33%) patients, overall in three right-handers (17%) who showed net and congruent leftward trunk and PV tilts. Two of the 18 (11%) had an outstanding clinical picture associating leftward: predominant parkinsonian symptoms, whole-body tilt (head -11°, trunk -8°) and transmodal tilt in verticality perception (PV -10°, VV -8.9°). Trunk orientation or VV were not modulated by STN-DBS, whereas PV tilts were attenuated by unilateral or bilateral stimulations if it was applied on the opposite STN. In most cases of PS, verticality perception is altered by the body deformity. In some cases, PS seems secondary to a biased internal model of verticality, and DBS on the side of the most denervated STN attenuated PV tilts with a quasi-immediate effect. This is an interesting track for further clinical studies.

Surface EMG vs. High-Density EMG: Tradeoff Between Performance and Usability for Head Orientation Prediction in VR Application

Head orientation prediction is one of the solutions to reduce end-to-end latency on Virtual Reality (VR) systems and is important since it can alleviate negative effects like motion sickness. This study compared head orientation prediction models from two different electromyography (EMG) systems: surface EMG (sEMG) and High-Density EMG (HD-EMG). The deep learning method was used to train the prediction model, and the results showed that the model with input from the pre-processed sEMG + IMU sensor outperformed the model with input from the HD-EMG + IMU sensor. However, the decreasing performance from HD-EMG was compensated by its comfort and the ease of use of its electrode. This tradeoff between performance and usability with sEMG compared to HD-EMG should be a consideration for users who want to choose between performance and ease of use for head orientation prediction purposes. Comparison with state-of-the-art head prediction methods proved that the sEMG-based model offers better performance in predictions when users change their head directions, which was quantified by calculating the dt peaks. In other words, our sEMG-based prediction model is suitable for VR applications, which require the user to perform high-intensity or abrupt movements, such as in FPS games or exercise/sports games.