Cross-domain Framework Research Articles

An improved deep learning approach for speech enhancement

Single-channel speech enhancement refers to the task of improving the quality and intelligibility of a speech signal in a noisy environment. Time-domain and time-frequency-domain methods are two main categories of approaches for speech enhancement. In this paper, we propose a approach based on a cross-domain framework. This framework utilizes our knowledge of the spectrogram and overcomes some of the limitations faced by time-frequency domain methods. First, we apply the intrinsic mode functions of the empirical mode decomposition and an improved version of principal component analysis. Then, we design a cross-domain learning framework to determine the correlations along the frequency and time axes. At low SNR = -5 dB, the effectiveness of our proposed approach is demonstrated by its performance based on objective and subjective measures. With average scores of -0.49, 2.47, 2.44, and 0.68 for SegSNR, PESQ, Cov, and STOI, respectively. The results highlight the success of our approach in addressing low SNR conditions.

Toward Source-Free Cross Tissues Histopathological Cell Segmentation via Target-Specific Finetuning.

Recognition and quantitative analytics of histopathological cells are the golden standard for diagnosing multiple cancers. Despite recent advances in deep learning techniques that have been widely investigated for the automated segmentation of various types of histopathological cells, the heavy dependency on specific histopathological image types with sufficient supervised annotations, as well as the limited access to clinical data in hospitals, still pose significant challenges in the application of computer-aided diagnosis in pathology. In this paper, we focus on the model generalization of cell segmentation towards cross-tissue histopathological images. Remarkably, a novel target-specific finetuning-based self-supervised domain adaptation framework is proposed to transfer the cell segmentation model to unlabeled target datasets, without access to source datasets and annotations. When performed on the target unlabeled histopathological image set, the proposed method only needs to tune very few parameters of the pre-trained model in a self-supervised manner. Considering the morphological properties of pathological cells, we introduce two constraint terms at both local and global levels into this framework to access more reliable predictions. The proposed cross-domain framework is validated on three different types of histopathological tissues, showing promising performance in self-supervised cell segmentation. Additionally, the whole framework can be further applied to clinical tools in pathology without accessing the original training image data. The code and dataset are released at: https://github.com/NeuronXJTU/SFDA-CellSeg.

Frequency Information Enhanced Deep EEG Denoising Network for Ocular Artifact Removal

Electroencephalography (EEG) signals are easily contaminated by various artifacts, making noise removal an essential step in EEG analysis. In recent years, deep-learning-based methods have provided a promising avenue for EEG denoising, with superior performance on several benchmarks. However, most existing deep models only rely on the temporal features of the signal and ignore the intrinsic characteristics in other domains, possibly limiting their ability to suppress artifacts. Therefore, we propose a cross-domain framework by integrating knowledge of both time and frequency domains into a deep-learning model for ocular artifact removal from EEG recordings. The proposed framework can be flexibly implemented on various deep denoising networks to improve their denoising performance. Extensive experiments demonstrate that the cross-domain design is capable of better eliminating the ocular artifact when implemented over several representative denoising networks. For example, RRMSE is reduced by 4%–26% for semisimulated data. These results verify the effectiveness of the proposed framework, which also indicates the importance of combining domain knowledge into the deep-learning models for EEG denoising.

A SONIFICATION OF THE ZCOSMOS GALAXY DATASET

The sonification of scientific data is useful not only for the work of visually impaired scientists but also for having a feeling of the global behaviour of a set of data and has also an intrinsic artistic value. Here we present the sonification of the zCOSMOS spectroscopic survey aimed to study the evolution of galaxies within the last 6 billions years of the Universe's life. The goals of such an initiative are multiple: providing a sound-based description of the dataset in order to make hidden features emerge, hybridizing science with art in a cross-domain framework, and treating scientific data as cultural heritage to be preserved and enhanced, thus breaking down the barriers between scientists and the general audience.

From “Land-Oriented” to “Human-Oriented”: Research on Evolution Features of China’s Industrial Park Planning Standards

This paper uses the time-series analysis function in CiteSpace software to systematically review the planning standards of Chinese industrial parks between 1985 and 2021, and then studies the framework, content, regions, and industry characteristics of these standards. The study shows that China’s industrial park planning standards have evolved through a series of developmental stages, namely from being land-oriented, then sector-oriented, and, most recently, human-oriented. The study also shows that through this process, a multi-layered, cross-domain framework of park planning standards has emerged. Due to the challenges posed by reforms in national land and spatial planning systems, the movement to transform and upgrade China’s industrial parks, as well as the rapid development of overseas parks, and the further development of industrial park planning standards, offer important new opportunities. Towards this end, this paper concludes by putting forward several strategies on how to best improve the framework of these standards, which is based upon keeping up with advanced technologies and basing development on practice.

Climate change, heat exposure and social impacts in and beyond the workplace: a systematic review and development of a cross domain framework for analysis and assessment

Climate change, heat exposure and social impacts in and beyond the workplace: a systematic review and development of a cross domain framework for analysis and assessment

A Framework for Development Architecture for Modular Products: Cross-Domain Variety Management Perspective

AbstractVariety management is a cross-domain issue in product family design. In the real field, the relationships across the domains are so complex for most of the existing product families that they cannot be easily identified without proper reference architecture. This reference architecture should provide the cross- domain mapping mechanisms in an explicit manner and be able to identify the proper units for management. From this perspective of cross-domain framework, this paper introduces development architecture (DA) to describe the relationships between elements in market, design, and production domains and to give insights for the cross-domain variety management in the product development stage. DA has three parts: (1) the arrangement of elements in each domain, (2) the mapping between elements, and (3) the identification of management sets and key interfaces which are the proper units for variety management. The proposed development architecture framework is applied to the case of front chassis family of modules of an automobile.

A cross-domain framework for designing healthcare mobile applications mining social networks to generate recommendations of training and nutrition planning

Nowadays, people are practicing physical exercise in order to maintain good health conditions. Such physical workouts are required by a plan, which should be designed and supervised by sport specialists and medical assistants. Thus, the exercise sessions shall start with consultation of a coach, doctor and dietician; however, many times this scenario is not presented. In typical activities such as running, cycling and fitness, people use health mobile apps with their smartphones, which offer support for these activities. Nevertheless, the functionality and operation of these applications are isolated, because many and long questionnaires are performed. Additionally, the physical and health state of a user is not considered. These issues would be taken into account for determining recommendations about the time for doing exercise and the kind of activity for each person. In this work, a social semantic mobile framework to generate recommendations where a mobile application allows sensing the physical performance, taking into consideration medical criteria with smartphones is proposed. The approach includes a semantic cross-information that comes from social network and official data as well as sport activities and medical knowledge. This knowledge is translated into application ontologies related directly to health, nutrition and training domains. The methodology also covers physical fitness tests and a monitoring tool for evaluating the nutrition plan and the correct execution of the training. As case study, the mobile application offers to evaluate the physical and health conditions of a runner, automatically generate a nutrition plan and training, monitor plans and recomputed them if users make changes in their routines. The data provided from the social network are used as feedback in the application, in order to make the training and nutrition plans more flexible by applying spatio-temporal analysis based on machine learning. Finally, the generated training and nutrition plans were validated by specialists, they have demonstrated 82% of effectiveness rate in exercise training routines and 86% in nutrition plans. In addition, the results were compared with isolated approaches and manual recommendations made by specialists, the obtained overall performance was 81%.

Describing complex design practices with a cross-domain framework: learning from Synthetic Biology and Swarm Robotics

This paper reports on the development of a cross-domain framework for describing complex design practices. The framework is grounded in studies of two different complex design fields: Synthetic Biology and Swarm Robotics. In the first study, we interviewed practitioners in Synthetic Biology, identifying three essential aspects of complex design problems and practices. The first of these aspects is the characterisation of system complexity, the second is the design objective taken with respect to this complexity, and the third is the design approach applied to realise this objective. In the second study, we interviewed designers in Swarm Robotics, confirming the domain generality of the three aspects identified in the first study and permitting a comparison to be made of how the two fields differ from each other in these aspects. Considered together, the two studies provide the basis for building a cross-domain framework for describing complex design practices. Such a framework is presented here, not to exhaust all possible descriptions of complex design practice but rather to provide a structured yet adaptable way of highlighting the important aspects of these descriptions. Indeed, each aspect of complex design can be can be broken down into different elements depending on the design contexts under consideration. Having such a framework enables designers to identify fundamental similarities and differences both between and within fields.

Situation awareness systems, states and processes: a holistic framework

This paper argues that state, system, and process descriptions of situation awareness (SA) are interdependent. Based on SA research from the last 30 years, the paper proposes a holistic SA framework. SA states emerge from processes of exploring situations through SA systems. Reflecting research on safety II (resilience), in describing SA states, the framework distinguishes frames (what situations are considered) from implications (regarding the situations) of objects on and of an event horizon. The paper describes and discusses SA system and process dependencies on SA states. It also describes SA system components as mediators and catalysts for SA, SA system properties (e.g. buffering SA), and dynamic SA system formation. Based on an analysis of four enactments of Air Traffic Control situations, the paper argues that what is domain-characteristic may not characterise all situations in a domain. The SA field could thus benefit from incrementally refining a nuanced cross-domain framework.

TCloud: A Dynamic Framework and Policies for Access Control across Multiple Domains in Cloud Computing

In a cloud computing environment, access control policy is an effective means of fortification cloud users and cloud resources services against security infringements. Based on analysis of current cloud computing security characteristics, the preamble of the concept of trust, role-based access control policy, combined with the characteristics of the cloud computing environment, there are multiple security management domains, so a new cross domain framework is for access control is proposed which is based on trust. It will establish and calculate the degree of trust in the single as well as multiple domains. Role Based Access Control is used for the implementation of the access control policies in a single domain environment with the introduction of the trust concept. In multiple domains the access control will be based on the conversion of roles. On the basis of trust, and role based access control model, a new novel framework of flexible cross domain access control framework is presented. The role assignment and conversion will take place dynamically.