Log in
Research in robust speech enhancement and audio-visual processing has led to impact on a range of different fronts:
(i) Collaboration with CSR, a leading $1 billion consumer electronics company, has shaped its R&D research agenda in speech enhancement, has inspired ideas for new product improvements, and has helped establish Belfast as an audio research centre of excellence within the company.
(ii) Our technology has changed the strategic R&D direction of a company delivering healthcare monitoring systems, with potential for multi-million pound savings in NHS budgets.
(iii) Audio-visual speech processing research has led to a proof-of-concept biometric system, Liopa: a novel, robust and convenient person authentication and verification technology exploiting lip and facial movements (www.liopa.co.uk). A start-up company is in an advanced stage of being established to commercialise this product. The product and commercialisation strategy was awarded First Prize in the 2013 NISP Connect £25K entrepreneurship competition in the Digital Media and Software category. The first commercial partner for Liopa has been engaged.
(iv) A system-on-chip implementation of a version of our speech recognition engine, which was developed through an EPSRC project, was awarded first prize in the High Technology Award in the 2010 NISP £25K Awards competition, and contributed to the founding of a spin-out company, Analytics Engines (www.analyticsengines.com).
Speech Graphics Ltd is a spinout company from the University of Edinburgh, building on research into the animation of talking heads during 2006-2011. Speech Graphics' technology is the first high fidelity lip-sync solution driven by audio. Speech Graphics market a multi-lingual, scalable solution to audio-driven animation that uses acoustic analysis and muscle dynamics to drive the faces of computer game characters accurately matching the words and emotion in the audio. The industry-leading technology developed by Speech Graphics has been used to animate characters in computer games developed by Supermassive games in 2012 and in music videos for artists such as Kanye West in 2013.
This impact case study provides evidence of economic impacts of our research because:
i) a spin-out company, Speech Graphics Ltd, has been created, established its viability, and gained international recognition;
ii) the computer games industry and the music video industry have adopted a new technology founded on University of Edinburgh research into a novel technique to synthesize lip motion trajectories using Trajectory Hidden Markov Models; and
iii) this led to the improvement of the process of cost-effective creation of computer games which can be sold worldwide because their dialogue can be more easily specialised into different human languages with rapid creation of high-quality facial animation replacing a combination of motion capture and manual animation.
Forensic speaker comparison is the analysis of recorded speech with evidential value in legal (usually criminal) cases. It is now routinely undertaken in the UK (ca. 600 cases annually) and increasingly elsewhere. It is vital that casework is underpinned by robust research, that reliable methods are applied, and that evidential results are framed appropriately. York is one of the world's largest research groups in forensic speech science, and in those academic disciplines (phonetics, sociolinguistics, sociophonetics) that provide the essential foundation for this applied field. The impacts of York research are felt through (i) enhancing understanding of variation in speech; (ii) applying research findings via collaboration in casework and research with J P French Associates (JPFA), one of the world's leading laboratories; (iii) providing doctoral research supervision for JPFA staff and professional training for other experts; (iv) providing expert evidence in legal cases in the UK and internationally; and (v) improving policy on expert evidence in the UK.
Increasingly in court cases the recorded voice of a perpetrator has to be compared with that of a suspect. Research on speaker characteristics carried out by/under Prof. Nolan has directly contributed to the work of those offering forensic speech services commercially or developing relevant speech processing software. Impact arises from seminal ideas such as LTF (Long Term Formant) analysis, and from the 100-speaker `DyViS' accent-matched database. The latter has directly enabled: the testing of an automatic speaker recognition system preparatory to its incorporation into forensic casework; the development of speaker recognition and speaker separation software; the adoption of systematic `voice quality' analysis; and the availability for casework of population statistics on pitch and disfluencies. Public engagement has raised awareness of the possibilities and limitations of speaker identification in legal and general audiences.
Stroke and other forms of brain injury often result in debilitating communication impairments. For example, patients with acquired apraxia of speech (AOS) experience difficulties that affect their capacity to verbally express thoughts and needs. Such individuals have benefitted from the development of a novel computerised treatment — "Sheffield Word" (SWORD). Patients who took part in clinical trials showed improvements in aspects of speech that were impaired after stroke. SWORD is now used by healthcare teams worldwide, providing benefits to a large patient population. The SWORD computerised treatment is convenient to use at home, fosters users' autonomy, and delivers higher treatment doses than possible through traditional clinical sessions. Clinicians who treat AOS have also benefitted through education, training and access to online materials about SWORD which were provided by the research team.
Our research on speech synthesis is embodied in software tools which we make freely available. This has led to widespread use and commercial success, including direct spinouts, follow-on companies and use by major corporations. This same research benefits people who lose the ability to speak and have to rely on computer-based communication aids. Unlike existing aids, which provide a small range of inappropriate voices which are often not accepted by users, our technology can uniquely create intelligible and normal-sounding personalised voices from recordings even of disordered speech, and so enable people to communicate and retain personal identity and dignity.
Nearly every large-vocabulary speech recognition system in current use employs outputs from fundamental research carried out in the University of Cambridge Department of Engineering (DoEng) on adaptation of Hidden Markov Models (HMMs). One example of the commercial application of these outputs is their use on the Microsoft Windows desktop for both the command and control functions and the dictation functions. Approximately one billion copies of Windows have been shipped since 2008. Other examples show the outputs used in the automatic transcription of a wide range of types of data. [text removed for publication]
One of the world-leading systems for large-vocabulary Automatic Speech Recognition (ASR) has been developed by a team led from the University of Sheffield. This system, which won the international evaluation campaigns for rich speech transcription organised by the US National Institute for Standards and Technology (NIST) in 2007 and 2009, has led directly to the creation of one spin-out, been largely instrumental in the launch of a second, has had significant impact on the development and growth of three existing companies, and has made highly advanced technology available free for the first time to a broad range of individual and organisational users, with applications including language learning, speech-to-speech translation and access to education for those with reading and writing difficulties.
Articulate Instruments Ltd. was founded in 2003 as a research, design, manufacturing and consulting company for users of phonetic instrumentation. It invents, designs, markets and supports instrumental technologies for normative and clinical speech science and for the diagnosis and treatment of speech disorders. Products include electronic systems, headsets, software, and methodologies, underpinned by QMU research. Clinical use of relevant products as medical devices requires "CE marking" to prove on-going safety and support, first achieved in 2004.
Impact relates primarily to the company's on-going financial health and its non-academic customer base. In its first 10 years, turnover averaged ~£120k, with over 200 customers internationally, of whom more than 50 were non-academic.
The impact is primarily in Public Health. It mainly concerns the adoption of and demand for a speech research technology, Electropalatography (EPG), for clinical diagnosis and treatment of speech disorders. Our continuing long-term and interdisciplinary research into EPG has increased our impact in this census period from the previous RAE2008, during which time the UOA had already been awarded a Queen's Anniversary Prize (2002) for working towards the clinical application of speech science.
Financial Support from the charitable sector and the NHS for the training of classroom assistants and SLTs in EPG therapy is highlighted, along with user testimonials, unmet demand, and small-scale provision of the therapy.