FreeGaze: A Framework for 3D Gaze Estimation Using Appearance Cues from a Facial Video.

Gaze is a significant behavioral characteristic that can be used to reflect a person's attention. In recent years, there has been a growing interest in estimating gaze from facial videos. However, gaze estimation remains a challenging problem due to variations in appearance and head poses. To address this, a framework for 3D gaze estimation using appearance cues is developed in this study. The framework begins with an end-to-end approach to detect facial landmarks. Subsequently, we employ a normalization method and improve the normalization method using orthogonal matrices and conduct comparative experiments to prove that the improved normalization method has a higher accuracy and a lower computational time in gaze estimation. Finally, we introduce a dual-branch convolutional neural network, named FG-Net, which processes the normalized images and extracts eye and face features through two branches. The extracted multi-features are then integrated and input into a fully connected layer to estimate the 3D gaze vectors. To evaluate the performance of our approach, we conduct ten-fold cross-validation experiments on two public datasets, namely MPIIGaze and EyeDiap, achieving remarkable accuracies of 3.11° and 2.75°, respectively. The results demonstrate the high effectiveness of our proposed framework, showcasing its state-of-the-art performance in 3D gaze estimation.

Higher-order figure-8 microphones/hydrophones collocated as a perpendicular triad-Their "spatial-matched-filter" beam steering.

Directional sensors, if collocated but perpendicularly oriented among themselves, would facilitate signal processing to uncouple the azimuth-polar direction from the time-frequency dimension-in addition to the physical advantage of spatial compactness. One such acoustical sensing unit is the well-known "tri-axial velocity sensor" (also known as the "gradient sensor," the "velocity-sensor triad," the "acoustic vector sensor," and the "vector hydrophone"), which comprises three identical figure-8 sensors of the first directivity-order, collocated spatially but oriented perpendicularly of each other. The directivity of the figure-8 sensors is hypothetically raised to a higher order in this analytical investigation with an innocent hope to sharpen the overall triad's directionality and steerability. Against this wishful aspiration, this paper rigorously analyzes how the directivity-order would affect the triad's "spatial-matched-filter" beam's directional steering capability, revealing which directivity-order(s) would allow the beam-pattern of full maneuverability toward any azimuthal direction and which directivity-order(s) cannot.

Three-dimensional dislocations in a uniform linear array's isotropic sensors-Direction finding's hybrid Cramér-Rao bound.

The linear array's one-dimensional spatial geometry is simple but suffices for univariate direction finding, i.e., is adequate for the estimation of an incident source's direction-of-arrival relative to the linear array axis. However, this nominal one-dimensional ideality could be often physically compromised in the real world, as the constituent sensors may dislocate three-dimensionally from their nominal positions. For example, a towed array is subject to ocean-surface waves and to oceanic currents [Tichavsky and Wong (2004). IEEE Trans. Sign. Process. 52(1), 36-47]. This paper analyzes how a nominally linear array's one-dimensional direction-finding accuracy would be degraded by the three-dimensional random dislocation of the constituent sensors. This analysis derives the hybrid Cramér-Rao bound (HCRB) of the arrival-angle estimate in a closed form expressed in terms of the sensors' dislocation statistics. Surprisingly, the sensors' dislocation could improve and not necessarily degrade the HCRB, depending on the dislocation variances but also on the incident source's arrival angle and the signal-to-noise power ratio.

The Use of Serratus Anterior Fascial Flap in Integrated Mastectomy and Implant Reconstruction.

BACKGROUND: Tissue reinforcement with acellular dermal matrix (ADM) in implant-based breast reconstruction contributes to the coverage of the implant and avoids its direct exposure to skin incision, yet not without risk of infection. Our integrated technique makes use of the in situ serratus anterior fascia as a support of the implant following mastectomy, which serves the same purpose of ADM in terms of aesthetic outcomes, but minimizes the hazard of infective complications. METHODS: We retrospectively reviewed all the nipple-sparing mastectomies with direct-to-implant immediate reconstruction in Hong Kong Sanatorium and Hospital from 2012 to 2016. The authors made use of the serratus anterior fascial flap as inferolateral coverage for the subpectoral implant. Consequently, the implant would be completely covered by autologous tissues. RESULTS: Among the 51 women included, primary breast cancers account for 91.8% of our indications for these 61 procedures of integrated mastectomy and implant reconstruction. The remaining five (8.2%) were performed as contralateral prophylactic mastectomy. Almost three quarters of the patients had a bra cup size of B or below. After a mean follow-up of 28.9 months, there was no reported post-operative complication of skin flap or nipple-areolar complex necrosis, or infection or extrusion of the implant. CONCLUSIONS: Our series support that the serratus anterior fascial flap could provide autologous coverage in integrated mastectomy and implant breast reconstruction, especially in small- and medium-sized breasts. Appropriate patient selection, as well as meticulous surgical technique, is critical for its success.

Hybrid Cramér-Rao bound of direction finding, using a triad of cardioid sensors that are perpendicularly oriented and spatially collocated.

Cardioid microphones/hydrophones are highly directional acoustical sensors, which enjoy easy availability via numerous commercial vendors for professional use. Collocating three such cardioids in orthogonal orientation to each other, the resulting triad would be sharply directional yet physically compact, while decoupling the incident signal's time-frequency dimensions from its azimuth-elevation directional dimensions, thereby simplifying signal-processing computations. This paper studies such a cardioid triad's azimuth-elevation direction-of-arrival estimation accuracy, which is characterized here by the hybrid Cramér-Rao bound. This analysis allows the cardioidicity index (α) to be stochastically uncertain, applies to any cardioidic order (k), and is valid for any real-valued incident signal regardless of the signal's time-frequency structure.

Bivariate direction finding using two perpendicular bi-directional ("figure-8") sensors of (possibly) unequal orders.

A "figure-8" sensor is so labeled because its spatial pattern resembles the character "8" with regard to the sensor's axis. This figure-8 pattern narrows as the sensor's order increases. Using two such figure-8 directional sensors of higher order, oriented perpendicularly to each other-this paper pioneers closed-form signal-processing algorithms to estimate an incident signal's azimuth-elevation bivariate direction-of-arrival. Monte Carlo simulations verify these proposed algorithms' efficacy and statistical closeness to the corresponding Cramér-Rao bounds.

Bilateral tibial tubercle avulsion fractures: A pediatric orthopedic injury at high risk for compartment syndrome.

Adolescent tibial tubercle avulsion fractures represent an uncommon, but clinically significant condition for emergency medicine physicians. Early recognition of the signs and symptoms of this pediatric orthopedic diagnosis are important, as anterior compartment syndrome can occur in up to 10-20% of cases. Anterior tibial tubercle fractures are generally sport related injuries, occurring primarily in otherwise healthy adolescent males between the ages of 11-17. They account for less than 3% of all epiphyseal injuries in this age group and are rarely bilateral in nature. In this article, we present a case with two unique clinical features: bilateral sports related tibial tubercle avulsion fractures and subsequent development of bilateral clinical compartment syndrome. We briefly review the risk factors, presentation, and diagnosis of this rare but clinically important condition.

Cardioid microphones/hydrophones in a collocated and orthogonal triad-A steerable beamformer with no beam-pointing error.

Cardioid sensors offer low sidelobes/backlobes compared to figure-8 bi-directional sensors (like velocity-sensors). Three cardioid sensors, in orthogonal orientation and in spatial collocation, have recently been proposed by Wong, Nnonyelu, and Wu [(2018). IEEE Trans. Sign. Process. 66(4), 895-906] and such a cardioid-triad's "spatial matched filter" beam-pattern has been analyzed therein. That beam-pattern, unfortunately, suffers pointing error, i.e., the spatial beam's actual peak direction deviates from the nominal "look direction." Instead, this paper will propose a steerable data-independent beamformer for the above-mentioned cardioidic triad to avoid beam-pointing error. Also analytically derived here (via multivariate calculus) is this beam-pattern's lobes' height ratio, beamwidth, directivity, and array gain.

A higher-order "figure-8" sensor and an isotropic sensor-For azimuth-elevation bivariate direction finding.

A "p-u probe" (also known as a "p-v probe") comprises one pressure-sensor (which is isotropic) and one uni-axial particle-velocity sensor (which has a "figure-8" bi-directional spatial directivity). This p-u probe may be generalized, by allowing the figure-8 bi-directional sensor to have a higher order of directivity. This higher-order p-u probe has not previously been investigated anywhere in the open literature (to the best knowledge of the present authors). For such a sensing system, this paper is first (1) to develop closed-form eigen-based signal-processing algorithms for azimuth-elevation direction finding; (2) to analytically derive the associated Cramér-Rao lower bounds (CRB), which are expressed explicitly in terms of the two constituent sensors' spatial geometry and in terms of the figure-8 sensor's directivity order; (3) to verify (via Monte Carlo simulations) the proposed direction-of-arrival estimators' efficacy and closeness to the respective CRB. Here, the higher-order p-u probe's two constituent sensors may be spatially displaced.

Near-field/far-field array manifold of an acoustic vector-sensor near a reflecting boundary.

The acoustic vector-sensor (a.k.a. the vector hydrophone) is a practical and versatile sound-measurement device, with applications in-room, open-air, or underwater. It consists of three identical uni-axial velocity-sensors in orthogonal orientations, plus a pressure-sensor-all in spatial collocation. Its far-field array manifold [Nehorai and Paldi (1994). IEEE Trans. Signal Process. 42, 2481-2491; Hawkes and Nehorai (2000). IEEE Trans. Signal Process. 48, 2981-2993] has been introduced into the technical field of signal processing about 2 decades ago, and many direction-finding algorithms have since been developed for this acoustic vector-sensor. The above array manifold is subsequently generalized for outside the far field in Wu, Wong, and Lau [(2010). IEEE Trans. Signal Process. 58, 3946-3951], but only if no reflection-boundary is to lie near the acoustic vector-sensor. As for the near-boundary array manifold for the general case of an emitter in the geometric near field, the far field, or anywhere in between-this paper derives and presents that array manifold in terms of signal-processing mathematics. Also derived here is the corresponding Cramér-Rao bound for azimuth-elevation-distance localization of an incident emitter, with the reflected wave shown to play a critical role on account of its constructive or destructive summation with the line-of-sight wave. The implications on source localization are explored, especially with respect to measurement model mismatch in maximum-likelihood direction finding and with regard to the spatial resolution between coexisting emitters.

A directionally tunable but frequency-invariant beamformer on an acoustic velocity-sensor triad to enhance speech perception.

Herein investigated are computationally simple microphone-array beamformers that are independent of the frequency-spectra of all signals, all interference, and all noises. These beamformers allow the listener to tune the desired azimuth-elevation "look direction." No prior information is needed of the interference. These beamformers deploy a physically compact triad of three collocated but orthogonally oriented velocity sensors. These proposed schemes' efficacy is verified by a jury test, using simulated data constructed with Mandarin Chinese (a.k.a. Putonghua) speech samples. For example, a desired speech signal, originally at a very adverse signal-to-interference-and-noise power ratio (SINR) of -30 dB, may be processed to become fully intelligible to the jury.