Imbalanced data and label scarcity pose significant challenges to gearbox fault diagnosis, often limiting the accuracy and robustness of diagnostic models. To address these issues, this paper proposes an imbalanced fault diagnosis framework based on an unsupervised subdomain adaptation mechanism with a dual-branch network. The framework first constructs a dual-branch feature extraction structure consisting of a lightweight backbone network (ResNet-18) and a deep backbone network (ResNet-50), which are used to capture low-level local features and high-level semantic representations, respectively. After the last convolutional layer of each branch, a long short-term memory network and a multi-head self-attention mechanism are embedded to integrate temporal dependencies and spatial contextual information, thereby enhancing the discriminability and generalization ability of the extracted features. Furthermore, a dynamic simulation model is established based on the nonlinear dynamic characteristics of the gearbox system to enrich the feature distribution and improve subdomain adaptability. Experimental results on multiple real-world gearbox fault datasets demonstrate that the proposed method achieves superior performance in terms of diagnostic accuracy and domain adaptability compared with existing approaches.

With the development of children’s education, parents and teachers pay more and more attention to the learning and training of children with developmental dyslexia (DD). There are many reasons for DD and various intervention tools with a large number of learning costs. In this paper, we aim to analyze the relationship between music, pinyin, and reading, and confirm the theoretical link of "music - Chinese pinyin - reading ability". Then, we sort out a method to design the music intervention tool in this article, which is a new tangible interactive tool for children with DD. This tool is composed of hardware and software and has three modules: music intervention, pinyin intervention, and reading intervention, covering a total of 11 levels and is suitable for classroom teaching, home self-study, and other scenarios, to improve the Chinese pinyin and reading abilities. Subsequently, we conducted a pre-experiment and formal experiment on 20 children with DD to demonstrate the effectiveness of this tool, while using the SUS scale to analyze the evaluation of these children. Finally, we figured that the intervention tool designed in this study has a certain helpful effect on improving the phonetic and reading abilities of children with DD, and is more effective than the three intervention methods used alone. This study has certain reference significance for the designers of intervention tools and also provides new intervention methods for teachers in training institutions. It also provides another possibility for the self-treatment and training of children with developmental dyslexia.

As a particular type of artificial intelligence technology, machine learning (ML) is widely used to empower user experience (UX). However, designers, especially the novice designers, struggle to integrate ML into familiar design activities because of its ever-changing and growable nature. This paper proposes a design method called Material Lifecycle Thinking (MLT) that considers ML as a design material with its own lifecycle. MLT encourages designers to regard ML, users, and scenarios as three co-creators who cooperate in creating ML-empowered UX. We have developed ML Lifecycle Canvas (Canvas), a conceptual design tool that incorporates visual representations of the co-creators and ML lifecycle. Canvas guides designers to organize essential information for the application of MLT. By involving design students in the “research through design” process, the development of Canvas was iterated through its application to design projects. MLT and Canvas have been evaluated in design workshops, with completed proposals and evaluation results demonstrating that our work is a solid step forward in bridging the gap between UX and ML.

A novel blind logo-watermarking algorithm for copyright protection of JPEG-compressed images is proposed. A visually meaningful grayscale logo is encoded by the (255, 9) BCH code into a codeword that is embedded as the watermark into the wavelet domain of the host image using a pixel-wise masking model. A trellis is generated with a specific path corresponding to that codeword. At the receiving end where the watermarked image is stored in a JPEG file, the codeword is recovered approximately through running the viterbi decoder on the trellis even without reference to the original host data, and then decoded back to the embedded logo by the (255, 9) BCH codec. Thanks to the BCH code’s powerful error correction capability and human vision system’s great tolerance to image noise, the extracted logo, possibly degraded, contains much more convincing information than that of a randomly-generated numerical sequence that is widely used in traditional watermarking strategies. Experimental results of successfully hiding a 32 × 32, grayscale logo into 512 × 512 JPEG-compressed images verify our algorithm’s practical performance.

Computer users vary greatly in their abilities to use a software interface efficiently. One factor that apparently affects users’ efficiency in using an interface is the changes in their skill levels. In this study, an adaptive interface (of menu and command) is presented that dynamically adjusts to users’ changing skill levels. The mechanism of an adaptive interface is described and discussed. The validity and usability of the adaptive interface is tested with 40 participants in an experiment that used a between‐subject experimental design for interface style. The independent variables were interface style (menu, command, hybrid, and adaptive) and skill level (starting session and ending sessions). The dependent variables were task completion time, number of steps used, ratio (of using menu mode over menu and command modes), perceived memory load, and satisfaction with the interface styles. The task‐completion time and ratio data indicate that the adaptive interface produced significantly better performance than the static hybrid interface at the end of the training sessions for experienced computer users. No significant differences were found for memory load and satisfaction ratings across the four interface styles.

Jirotka, M., and Goguen, J. A. (1994). Requirements Engineering: Social and Technical Issues. 296 pages, US$52.50. San Diego: Academic Press, Inc. (ISBN 0–1238–5335–4) Dillon, A. (1994). Designing Usable Electronic Text: Ergonomie Aspects of Human Information Usage. 175 pages, Cloth $75.00, Paper $32.00. Bristol, PA: Taylor & Francis, Ltd. (ISBN 0–7484–0113‐X) Oppermann, R. (1994). Adaptive User Support. 253 pages, US$49.95. Hillsdale, NJ: Lawrence Erlbaum Associates, Inc. (ISBN 0–8058–1655–0) Landauer, T. K. (1995). The Trouble with Computers. 296 pages, US$27.50. Cambridge, MA: The MIT Press. (ISBN 0–262–12186–7). Weimer, J. (1995). Research Techniques In Human Engineering. 504 pages, US$63.75. Englewood Cliffs, NJ: Prentice Hall PTR. (ISBN 0–13–097072–7).