Queen Mary, University of London CSC Scholarship PhD project
24741
伦敦大学玛丽女王学院（Queen Mary, University of London）EECS学院CSC奖学金PhD项目目前还有一个招生名额，入学时间为今年9月，PhD项目学制4年，QM免去所有学费，CSC（国家奖学金）承担所有生活费。
语言成绩要求：IELTS 6.5或以上，TOEFL 95或以上
以下是具体project的介绍：
Supervisor: Prof. Mark Plumbley (Director of Centre for Digital Music)
elec.qmul.ac.uk
Dr. Dawn Black
eecs.qmul.ac.uk
Title: Tagging consumer multimedia content using audio features
An increasing number of people are now using video and audio channels such as YouTube and SoundCloud to upload their recorded content to the web. However, much of this information is poorly tagged, meaning that it can be hard for other people to find relevant content.
The aim of this project is to investigate the use the audio information only to assign meaningful tags to this type of content, and so allow content of interest to be more easily discovered. Recent research in this area has demonstrated that this is feasible, confirming the informal idea that it is possible to guess the topic of a video just by listening to the sound. For example, a sports match sounds very different to a birthday party.
However, many videos correspond to multiple concepts, so these need multiple tags. To tackle this, it is proposed to use methods such as "sparse representations", "non-negative matrix factorization" (NMF) and "deep learning" to learn multiple features from audio data, tailored to this task through dictionary learning and other feature learning methods. We will investigate the types of representations that emerge from these approaches, and compare to state of the art methods.
QM的数字音乐中心（C4DM）是全欧领先的音乐及音频技术研究组，成立于2003年，目前有超过55名研究人员，每年研究经费超过1700万英镑，具备先进的研究设备。实验室主任Mark Plumbley教授长期从事独立成分分析（ICA）及稀疏表达（sparse representation）的研究，经验丰富，而且去年达到了教学时间要求，将不再担任授课工作，可专心带领PhD进行研究。
如有任何疑问，请随时与我联系（qq: 396977035; Email: 1point3acres.com）
如感兴趣，可以直接将CV发至我的邮箱，将尽快安排面试。
语言成绩要求：IELTS 6.5或以上，TOEFL 95或以上
以下是具体project的介绍：
Supervisor: Prof. Mark Plumbley (Director of Centre for Digital Music)
elec.qmul.ac.uk
Dr. Dawn Black
eecs.qmul.ac.uk
Title: Tagging consumer multimedia content using audio features
An increasing number of people are now using video and audio channels such as YouTube and SoundCloud to upload their recorded content to the web. However, much of this information is poorly tagged, meaning that it can be hard for other people to find relevant content.
The aim of this project is to investigate the use the audio information only to assign meaningful tags to this type of content, and so allow content of interest to be more easily discovered. Recent research in this area has demonstrated that this is feasible, confirming the informal idea that it is possible to guess the topic of a video just by listening to the sound. For example, a sports match sounds very different to a birthday party.
However, many videos correspond to multiple concepts, so these need multiple tags. To tackle this, it is proposed to use methods such as "sparse representations", "non-negative matrix factorization" (NMF) and "deep learning" to learn multiple features from audio data, tailored to this task through dictionary learning and other feature learning methods. We will investigate the types of representations that emerge from these approaches, and compare to state of the art methods.
QM的数字音乐中心（C4DM）是全欧领先的音乐及音频技术研究组，成立于2003年，目前有超过55名研究人员，每年研究经费超过1700万英镑，具备先进的研究设备。实验室主任Mark Plumbley教授长期从事独立成分分析（ICA）及稀疏表达（sparse representation）的研究，经验丰富，而且去年达到了教学时间要求，将不再担任授课工作，可专心带领PhD进行研究。
如有任何疑问，请随时与我联系（qq: 396977035; Email: 1point3acres.com）
如感兴趣，可以直接将CV发至我的邮箱，将尽快安排面试。
- 1
1条回复