Recognizing facial expression has attracted much more attention due to its broad range of applications in human–computer interaction systems. Although facial representation is crucial to final recognition accuracy, traditional handcrafted representations only reflect shallow characteristics and it is uncertain whether the convolutional layer can extract better ones. In addition, the policy that weights are shared across a whole image is improper for structured face images. To overcome such limitations, a novel method based on patches of interest, the Patch Attention Layer (PAL) of embedding handcrafted features, is proposed to learn the local shallow facial features of each patch on face images. Firstly, a handcrafted feature, Gabor surface feature (GSF), is extracted by convolving the input face image with a set of predefined Gabor filters. Secondly, the generated feature is segmented as nonoverlapped patches that can capture local shallow features by the strategy of using different local patches with different filters. Then, the weighted shallow features are fed into the remaining convolutional layers to capture high-level features. Our method can be carried out directly on a static image without facial landmark information, and the preprocessing step is very simple. Experiments on four databases show that our method achieved very competitive performance (Extended Cohn–Kanade database (CK+): 98.93%; Oulu-CASIA: 97.57%; Japanese Female Facial Expressions database (JAFFE): 93.38%; and RAF-DB: 86.8%) compared to other state-of-the-art methods.

In addition to high compliance to unstructured environments, soft robots can be further improved to gain the advantages of rigid robots by increasing stiffness. Indeed, realizing the adjustable stiffness of soft continuum robots can provide safer interactions with objects and greatly expand their application range. To address the above situation, we propose a tubular stiffening segment based on layer jamming. It can temporarily increase the stiffness of the soft robot in a desired configuration. Furthermore, we also present a spine-inspired soft robot that can provide support in tubular segments to prevent buckling. Theoretical analysis was conducted to predict the stiffness variation of the robot at different vacuum levels. Finally, we integrated the spine-inspired soft robot and tubular stiffening segment to obtain the tuneable-stiffness soft continuum robot (TSCR). Experimental tests were performed to evaluate the robot’s shape control and stiffness tuning effectiveness. Experimental results showed that the bending stiffness of the initial TSCR increased by more than 15× at 0°, 30× at 90°, and 60× in compressive stiffness.

Since climbing robots mainly rely on adhesion actuators to achieve adhesion, robust adhesion actuators have always been the challenge of climbing robot design. A novel under-actuated soft adhesion actuator (USAA) proposed in this paper for climbing robots can generate adhesion through robot’s load applied to the actuator. The actuator is composed of a soft film/substrate structure with an annular groove on the substrate and a cavity on the soft film. To fabricate the actuator, we first study the influence of the geometric parameters of the USAA on the maximum adhesion of the actuator by analysis and experiments, and then combine these parameters and the boundary conditions of the static meniscus in the mold to design the mold. Moreover, we fabricate a climbing robot equipped with USAAs and evaluate its performance on horizontal and inclined surfaces with a wide range of characteristics. The USAA can generate strong and controllable adhesion to various smooth and semi-smooth surfaces. Furthermore, the fabricated robot performs well on various surfaces under a certain load (at least 500 g) and speed (369 mm/min) through experiments. It’s adaptability to a variety of surfaces enables a wide range of applications and pushes the boundaries of soft adhesion actuators.

Speech emotion recognition (SER) has become an attractive topic owing to its broad range of applications. Segmentation is often used to increase training data for SER, but the inherited label may result in low performance. In this paper, we proposed a robust noise-label-suppressed module by relabeling the segment label to suppress the bad effects of the inherited label. Firstly, the segment of the log Mel spectrogram with deltas and delta-deltas of speech was calculated. Then, speech features were extracted by feature extraction model with 3-D data. Finally, the labels of each segment were corrected by the relabel model. Experimental results on the IEMOCAP dataset illustrate that our proposed noise-label suppressed module is superior to other advanced methods and gets robust performance.

Polyetheretherketone (PEEK) has shown broad application prospects in implantable medical devices, including orthopedic, craniofacial, spinal fusion, dental, and extravascular stent applications, owing to its elastic modulus comparable to human cortical bone, excellent mechanical properties, good biocompatibility, and radiolucency. However, the inherent bioinertness and hydrophobicity of PEEK hinder firm osseointegration with surrounding bone tissue after implantation and limit its multiple biological functions, such as antibacterial, anti-inflammatory, pro-angiogenic, and lubricating activities, thereby restricting its clinical use in complex physiological environments such as infection, post-tumor resection rehabilitation, and artificial joints. In recent years, inspired by natural bone tissue, extracellular matrix, biomineralization processes, and mussel adhesive proteins, biomimetic design strategies have been widely explored to enhance the bioactivity and mechanical adaptability of PEEK implants. This review systematically summarizes recent advances in the biomimetic design of PEEK implantable medical devices, focusing on two main aspects: structural biomimicry (porous/gradient structures, biomimetic lattices and layered architectures, and surface topographies) and surface functionalization (bioactive coatings and mineralization modification, and biomolecule/drug immobilization and controlled release). Their current applications in craniofacial reconstruction, spinal fusion, joint replacement, long-bone prostheses, dental implants, and extravascular stents are also summarized. In addition, current limitations of biomimetic design are analyzed, including coating stability, multifunctional integration, biofunctionalization of three-dimensional structures, and clinical translation. Future research should prioritize multifunctional synergy and spatiotemporal sequential regulation, smart microenvironment-responsive systems, precise fabrication of cross-scale gradient structures, and artificial intelligence-driven personalized design, thereby driving the translation of PEEK implants from passive substitutes to active next-generation high-performance medical devices capable of regulating the tissue microenvironment.