MScCT+ Students Won Awards at Silicon Valley International Invention Festival

The Intelligent Integrated Carbon Tracking Dogs and Carbon Removal Robots Deployment System uses highly sensitive gas sensors to continuously detect carbon dioxide concentrations in the environment, and autonomously locate and identify areas with a higher concentration.

The Carbon Tracking Dogs and Carbon Removal Robots then work in tandem to scan the surroundings, using 3D laser radar technology to generate a very accurate 3D map, which autonomously calculates the best way to avoid obstacles and navigate safely to the target area. The system has high-efficiency solar panels for continuous power support for efficient and high-purity carbon capture.

This is a real-time speech recognition system for edge devices based on dynamic audio segmentation, which can be used for conference recordings and real-time translation. The system first records audio in real time through a PyAudio library and sets a buffer to receive the recorded audio.

Then the complete audio in the buffer is transmitted to the speech recognition model at fixed intervals by asynchronous processing, using a non-streaming model, Whisper, to recognise the transmitted audio, and convert it into text.

The IntelliAnalyzer is a multimodal and multi-level industrial decision-making engine, which automatically analyses design schemes through natural language processing, image recognition, and knowledge reasoning. It identifies key process features automatically by the multimodal analysis of graphic and text design drafts, recommends the best design and production routes, dynamically calculates production costs, generates process optimisation plans and cost risk warnings, and accumulates business experience to form transferable knowledge assets.

This project focuses on the pain points of design-production-cost collaboration in the manufacturing industry, providing SaaS services to help enterprises reduce costs, increase efficiency, and achieve intelligent transformation to cope with marketisation. It solves problems that traditional analysis methods cannot handle, such as complex designs and dynamic bear market demands, covering high-end manufacturing such as furniture and medical equipment.

The system uses lightweight architecture and personalised expansion as its core, challenging three major contradictions: the mismatch between users' massive storage needs and the capacity of traditional devices, loss of storage efficiency caused by mixed functions, and usage barriers arising from complex operations for special groups.

A centralised + P2P hybrid architecture, compatible with older operating systems and low-configuration devices is adopted, which has no complex animation designs, reduces hardware usage, and improves running speed. Its lightweight core functions support on-demand plugins and extensions, minimising local resource consumption for basic functions, and enabling flexible adjustment implemented by third-party plugins. The system integrates three major accessible modules, providing eye control and expression recognition for people with limb disabilities, virtual keyboards, and voice assistants for those with visual impairments or limited mobility, and simplifying UI design to lower learning costs. It also fills the gap in dedicated storage for people with disabilities, requires no additional hardware investment, has low maintenance costs, supports stable operation in all scenarios, and achieves zero-threshold access and network-ready usage.

The system allows a choice of three different recommendation modes based on users’ preferences. It combines the strengths of content-based and collaborative filtering to provide personalised recommendations that adjust dynamically and allow flexible switching based on changes in the user's interests. A customised movie list is immediately generated by combining the user's personal information and past behaviours, and displays a tailored list of recommendations, which are always aligned with the user's current needs and preferences. The list is scalable, and suitable for integration across various industry platforms.

The system then allows users to comment on the recommendations list, records the feedback, and adjusts and optimises the algorithm to ensure that the recommendations suit the user.

The system is a clever combination of modern psychology, classic Chinese culture and fortune-telling, and cutting-edge AI technology. First it builds a three-dimensional data system of modern psychology, traditional numerology, and behavioural labelling, which overcomes the limitations of standard platforms that rely on basic information such as age and occupation for matching, and provides more in-depth and personalised services.

After collecting sufficient user information, the system quantifies it to form a unified data format to be analysed and processed by three innovative modules - the Multi-dimensional Data Quantification System, the Cross-domain Feature Fusion Engine, and the AI-driven Intelligent Interaction System - to optimise the matching algorithm and enhance accuracy.