A person's emotion can be represented through facial expressions in non-vocal communication. Nowadays, automatic facial expression recognition systems have attracted myriad interest in applications such as face biometric-based authentication, behavior analysis (psychology), health monitoring (cerebral palsy), recommendation systems, and many others. Deep learning-based solutions have become the most-handy method to solve any image-video processing problems in recent times. Nevertheless, these CNN models include many hidden layers with complex predefined mathematical functions, resulting in increased complexity. Therefore, deep architecture poses a challenge to deal with a large number of learning parameters. This manuscript proposes two customized-CNN models, named Proposed_Model_1 and Proposed_Model_2, to classify universal facial expressions without overfitting. In this paper, the effect of hyper-parameters such as activation function, learning rate, kernel size, and convolutional block are investigated and optimized efficiently. Experimental results reveal that both of our proposed models outperform other existing methods considering all universal facial expressions, with an accuracy of 67.24% and 66.61%, respectively, on the well-known benchmark public dataset (Facial Expression Recognition-2013).

An efficient CNN approach for facial expression recognition with some measures of overfitting

A photonic crystal based MOEMS pressure sensor design is presented in this work. The key design element is a circular diaphragm made of silicon (Si), integrated with a hexagonal ring resonator in the middle. The novelty of the work is the arrangement of the holes in the Si diaphragm in triangular orientation and optimization of sensitivity for pressure sensing. The pressure is applied at the center of the diaphragm which resulted in deformation, stress, and strain. The minimum observable deformation is 1.0073 nm and is at 6.3 µPa. The corresponding change in the refractive index is calculated for the stress applied. The resonant wavelength obtained for without load is 1.598 µm. The shift in the resonant wavelength for different loads are observed and tabulated. The modeling tool used to extract the mechanical parameters is ANSYS multiphysics. For 2D modeling and simulation of the ring resonator structure, MEEP is used for the implementation of FDTD. The Q-factor observed is 12,245 and sensitivity reported as 316.15 nm/RIU respectively.

Circular diaphragm-based MOEMS pressure sensor using ring resonator

The Bacillus cereus or B. cereus group comprises genetically closely related species with variable toxigenic characteristics. Most human food poisoning is caused by B. cereus. B. cereus is associated with nosocomial and opportunistic infections, particularly in immunocompromised patients. The motivation for this work was to provide an alternate method of detection of B. cereus on the fly. The purpose of the current study is to establish an innovative method to detect B. cereus. The proposed technique relies on a change in refractive index (RI) between normal and B. cereus infected blood. This involves the spectral analysis of the sample and uses the unique RI of the bacteria as a spectral signature for detection. The design incorporates an optical biosensor based on the photonic crystal (PC). Here we have considered two configurations—concentric and the side by side hexagonal ring resonators for PC based biosensor design. We have observed the shift in the frequency of the spectrum between normal and infected blood. From the plotted transmission spectrum, we conclude that there is a significant shift in frequency because of the change in RI due to infection. For the designed biosensors the ranges of Quality factors obtained are between 722 to 1626.31 and Sensitivity 1.0845 µm/RIU to 375.5 µm/RIU. This concludes that we can use B. cereus infected blood specific transmission spectra as a signature to match unknown blood sample to detect B. cereus.

Highly sensitive photonic crystal based biosensor for Bacillus cereus

In this work we proposed fibre Bragg grating (FBG) based pressure sensor, to detect the pressure in subaquatic application. For specific underwater pressure intervals, change in wavelength of FBG sensor is simulated with photonic design software. Obtained results exhibited distinct shift in peak resonant wavelength with wavelength ranging from 1550 to 1556 nm with Q factor of 1,32,280. FBG encapsulated mechanical model designed and analyzed to investigate strain behavior due to pressure application. Pressure and strain change values are overlapping for the operable wavelength of 1.5625 during mechanical and optical analysis. Absolute sensitivity of sensor found to be 0.04644/RIU. This proposed sensor can play significant roles in studying underwater diving conditions.

Grating based pressure monitoring system for subaquatic application

Theoretical investigation of Bragg Reflector optical sensor for the measurement of cryogenic temperature

Proposed work consist of simple and micro size optical MEMS based displacement sensor for detection of muscle movement in human body. Optical MEMS based structure, analyzed with rods-in air and hole-in slab configuration of photonic crystal. Remarkable shift in wavelength is observed in RIA configuration with high quality factor of 3245 compare to holes in slab in the photonic sensing structure with Q factor 1456. Optical sensitivity of device with rods in air configuration is 350 nm/RIU is obtained and from the result, the distinctly visible shift in wavelength is found to be optimum for fabrication of projected device. Proposed work is applicable for the muscle movement of human body during medical diagnosis and for daily life activities.

Design and analysis of moems based displacement sensor for detection of muscle activity in human body

Integrated MOEMS based cantilever sensor for early detection of cancer

Performance analysis of optomechanical‐based microcantilever sensor with various geometrical shapes

We explored the design of an optical pressure sensor. The objective is to create a photonic microelectromechanical system-based cantilever sensor design to detect prostate-specific antigen, a protein biomarker associated with prostate cancer. Sensor’s performance for the early detection of cancerous cells is dependent on sensitivity. The designed sensor consists of a hexagonal ring integrated with microcantilevers. Two types of microcantilever such as rectangular profile and V profile with integrated photonic sensing layer are investigated for sensitivity, quality factor, and resonant wavelength. The analysis shows that a geometrical modification of microcantilever has a significant effect on sensitivity enhancement. The finite difference time domain tool is used for designing photonic ring resonator patches. Numerical analysis of microcantilever is considered to investigate surface stress behavior. The cantilever deformation due to applied pressure resulted in a change in the index of refraction. Results show that the sensitivity of 55  nm  /  MPa for a triangular-shaped microcantilever obtained is higher compared to the rectangular-shaped microcantilever with a sensitivity of 0.19  nm  /  MPa. The designed structures have significantly higher sensitivities than the previously published sensitivity of 3.27  nm  /  MPa at wavelength 1550 nm. The maximum quality factor obtained for the rectangular shape is 2852 and 77,510 for the triangular shape. Triangular-shaped microcantilever can minimize winding inflexibility or stiffness. This capability of the triangular-shaped microcantilever enhances feasibility in making the final packaging and future fabrication.

Anup M. Upadhyaya

Papers

Design and analysis of moems based displacement sensor for detection of muscle activity in human body

Integrated MOEMS based cantilever sensor for early detection of cancer

Grating based pressure monitoring system for subaquatic application

Performance analysis of optomechanical‐based microcantilever sensor with various geometrical shapes

Silicon nanostructure-based photonic MEMS sensor for biosensing application