Machine learning & AI
Fully funded PhD Scholarship - Machine learning & AI for activity recognition (2019)
Type of award
== Objectives ==
AI and machine learning techniques can be used to infer human activities by interpreting the data originating from a variety of multimodal wearable, mobile and ambient sensors.
This can be used to provide contextual assistance with applications in human-robot interaction, healthcare, industrial assistance, sports training, skill assessment, entertainment, and others.
Within this project, you will seek methods to make it easier to train systems to recognise a wider range of human activities. You will research advanced machine learning and AI techniques to recognise a growing set of activities from multimodal sensors, and reduce the effort associated with acquiring annotated training data.
Depending on your interests, different approaches can be followed: deep transfer learning to exploit the growing availability of multimedia datasets (e.g. Google AVA dataset, Youtube data, the Sussex-Huawei dataset), interactive machine learning, crowd-sourcing, adaptive machine learning, and others.
The project can be oriented towards methods achieving high performance for offline usage, or methods suitable for real-time activity recognition running on embedded platorms. Demonstrators arising from this project are welcome.
== About the Lab ==
The Wearable Technologies Lab, led by Dr. Daniel Roggen, has been established in 2014. Since then, it has acquired funding from Google, Huawei, EPSRC, Unilever, the Austrian FFG, and others.
The focus of our lab is to advance AI techniques to automatically recognise and understand human activities or daily routines from wearable and mobile sensors. We have developed several wearable sensing platforms and software frameworks for this, including deep learning and ASIC-friendly approaches.
The lab has created numerous dataset for activity recognition research, the most recent is a massive transportation dataset - the Sussex-Huawei Locomotion dataset (www.shl-dataset.org) - which has been used in two prominent machine learning challenges at Ubicomp 2018 and Ubicomp 2019.
Some of our applications are in the fields of sports performance, industrial assistance, mobility monitoring, crowd behaviour analytics and healthcare.
The members of the lab have an international outlook, with a mix of computer scientists, computer engineers, and electronic engineers.
The lab has state of the art computing and electronics facilities with a wide range of technologies at hand: GPU computing platforms, augmented reality glasses, smartwatches, a vast array of datasets and ad-hoc software tools to support research, numerous novel sensor technologies and sensing platforms, etc.
£15,009 stipend, plus fees (at the UK/EU rate).
You will receive a tax free stipend at a standard rate of £15,009 per year for three years. In addition, your fees will be waived for three years (at the UK/EU rate).
Overseas applicants are welcome to apply if they can meet the fee shortfall.
Open to candidates from UK/EU countries and from non-EU countries.
The ideal candidates will have a master's degree in computer science, computer engineering, physics, mathematics, electrical engineering, or equivalent, with prior experience desired in machine learning and ideally in embedded and wearable systems. The candidate will have excellent technical skills, including programming in some of Python, C/C++ or Matlab and experience with deep learning frameworks and Linux.
The ideal candidates will have a passion to contribute to the development of novel wearables which can improve quality of life. They will have outstanding technical skills and a strong interest in research at the crossroads of signal processing, machine learning, embedded systems, sensor technologies and their applications.
Applicants should be committed to pursue leading research and publish results in top venues. Additionally, we expect mastery of written and spoken English, self-motivation, an inquiring mind, be able to work independently and in an interdisciplinary environment.
For more information click "LINK TO ORIGINAL" below.