Preface by Project Manager.
The TCALL project has now completed 27 months of development work. Our aim is to extend the use of telecommunications by the deaf. This includes examining the use of text- and pen- based graphics communication, FAX, video transmission of sign language and automatic translation between text and speech.
There is to be evaluation work with the deaf community and systems development based on this experience. We intend that our findings will influence the design of future telecommunications equipment so that the needs of the hearing disabled are met in production rather than being addressed as some later stage addition.
The five Partners are:-
University of Bristol - Centre for Communications Research.
Specifically providing research and system support expertise for network interface and terminal software design. ( CCR’s representative, Professor Andrew Bateman, left during the 1997 Autumn to pursue a career in industry. The TCALL Partners were sorry to loose his considered input, experience and stimulation. Mrs Dawn Bradford managed CCR’s interests in the project until January 1998 when Univ. of Bristol appointed Dr Michael Barton to the position.)
University of Bristol - Centre for Deaf Studies.
Specifically providing consumer research and the field trial programme.
University of Dundee - Centre for MicroElectronics.
Specifically interested in telephone applications for people without speech, and have interests in telecommunications for people with special needs generally.
Orange.
Provide mobile telephones and access to their telephone network for test purposes.(During 1997, Orange’s representative, Dr David Bozward left to pursue other interests on behalf of the company. TCALL welcomed Dr Martin Swinburne and Mr Kassir Hussain as his replacements. In January 1998, it was advised that Ms Anila Jacob would assume Dr Swinburne’s role in TCALL.)
Portset Systems Ltd.
Specifically interested in the Text-to-Speech and Speech-to-Text development work. Portset Systems Ltd is also the Project Manager.
TCALL funding is received from the D.T.I., from E.P.S.R.C. and from the industrial partners.
Non-Interactive Text Transfer
Prototype Laptop based portable SMS system and software support
It is possible to use SMS on a Laptop, using a Nokia phone and data card, with the Windows programs Terminal (WIN 3.1) and HyperTerminal (Windows 95). However the interface is unfriendly and to remedy this, a group of software packages, referred to as TextTalk, is being developed. This includes a package for sending a single SMS message. The main program allows sending, receiving, forwarding etc. of SMS messages, in addition to address book handling, logging and analysis of messages. The package is designed with the deaf specifically in mind, with commands and status being expressed in pictures as well as text. The address book also allows for identification by image as well as by name.
Prototype Laptop based interactive text system and software support using Data Service
The TextTalk package also provides a friendly interface to the Data service.
Further development during 1997 included implementation of a data answer phone. It had been found that it was difficult for deaf people to take data calls because it was necessary for the mobile phone and laptop to be connected in advance, and the TextTalk software to be in Data mode. This meant that if a call was received when the user was not very close to the equipment, they could be unaware of it and miss the call. The data answer phone answers data calls on the user’s behalf with a message (allowing call filtering) and after a delay sends another message to ask the caller to leave their message. Messages can be viewed and deleted on the user's return.
Facsimile Transmission
Several attempts have been made to use existing software to send faxes from a mobile phone, both to other mobiles and to static phones. Success rates were extremely low, and we understand this is due to hardware incompatibility. Until this problem is solved, it was not felt to be worthwhile incorporating facsimile transmission into the TextTalk package.
Graphics Transmission
A method of encoding, sending and receiving pictures in the TextTalk package is under development.
Enhanced Text Entry Software
Development of abbreviation expansion software
The TextTalk package allows expansion of abbreviated messages, and in addition allows compression of messages using recognised abbreviations.
Report on suitability of Grammar checkers
A review of commercially available grammar checking software was conducted with reference to its suitability for Deaf users.
User Trials
The Centre for Communications Research provided a help line for Phase 1 of the user trials. The experience gained from this was used to update and improve the TextTalk suite of Programs.
University of Bristol - Centre for Deaf Studies.
Although the technology now exists to provide telecommunications services to deaf people, there remain many questions concerning the interface and the use deaf people can make of the additional reach which the phone offers. The future for deaf telecommunications is in video communications with text options, nevertheless it is important to determine the social factors which lead to greater or lesser use of the phone by deaf people.
In TCALL, three successive field trials were conducted during 1997. These were similar in nature - five weeks, using SMS, individual and group tasks. Information from earlier trials influenced the running of the later trials and the groups chosen to participate were in increasing geographic distances from Bristol.
Although there have been hardware and software problems, the trials were completed with relatively few mishaps. The planned tests on mobile fax had to be abandoned due to an incompatibility between hardware and software.
Trials data was collected on the completion of tasks, on the participants and on their experience of the systems, on air time and networking, on message content, and on contact with the help desk. This data is complex and will not be completely analysed until February 1998. The preliminary findings include those below.
Data collected from the first trials provided a great deal of insight into the likely pattern of use of "committed" users. We could see major differences in the extent of use: Deaf people do not automatically take to the technology. The trials indicated the feasibility of SMS as a one-to-one communication medium. However, the same could not be said of SMS as a means of group exchange of information. A message "broadcast" to a group of people is not necessarily received by the individuals at the same time. If the phone is switched off, the message is not delivered at all if the SMS time limit is exceeded.
Use of the Notebook-PC with a data card and phone proved patchy for a number of reasons:
- The machine must be on continuously on in order to receive calls.
- It was difficult to set up the notebook especially if the phone was receiving a call.
- The physical size and weight of even the lightest Notebook-PC as currently available.
It seems that PC technology will need to become even more compact, or mobile phones will need to become capable of other functions, to stimulate interest.
The second trials were completed with revised, more stable TextTalk © software. The findings remain that SMS (in its most portable form) is of great value, but that working with the notebooks introduces complications which are less attractive to the volunteer user. The overall level of activity was reduced by about 20% compared to the first trials and was attributed to users being less well known to each other.
In trial three, the participants were much further afield and could be supported less well. The factors of not knowing each other and the distance from Bristol made the promises to participate weaken once the individual was at home. The level of calls dropped dramatically (50% down on Trials One) and a number of participants were unable to complete the tasks. Live text chat was mostly inoperable although a number were able to use it successfully. The impact of the trials was limited because only a third of those involved were able to claim a successful connection. Group work was again problematic and it is clear that the immediacy of hearing people's use of the phone is not replicated in the text-based systems we were able to offer.
During the trials, there was concern about the use of voice. The great majority of “voice” calls were in error and caused by pressing the call button on the phone. A Deaf user, beginning this process by error, is unable to detect the Orange answerphone and continues to hold as if the phone is ringing at the other end. Further consideration must be given to this.
Much data analysis remains to be done before a final report of these trials is written.
However, there is already a great deal of information upon which we can build services and applications and which will be vital to future planning.
University of Dundee - Centre for MicroElectronics.
There is an ongoing task to ensure that the TCALL developments are responsive to developments elsewhere in the Telecommunications and Data Communications fields.
The principle task this year has been an empirical comparison of various data communications and telecommunications technologies. This has met two objectives:
a) To allow equivalent fixed and mobile technologies to be compared.
b) To consider the types of data communications technologies that could be used to provide multimedia services.
It is feasible to consider services that allow users to exchange text phrases, picture files or video or audio clips alongside the text phone service being developed within TCALL.
The study compared mobile phone data services with the equivalent PSTN services, and also wireless LAN services with the equivalent fixed Ethernet LAN services. Wireless LAN technology is emerging as a higher bandwidth alternative to mobile or cordless phone links as a means of providing data services. The results of the study showed that, whilst the fixed link technologies gave superior raw data transfer performance, there was no measurable difference between the fixed and mobile technologies when the client machine was handling multiple tasks. This indicates that the link may not be the critical limiting factor, but that the ability of the computer platform to multitask, and to service it’s communications functions will probably limit the performance of the service.
Future work in this work package will concentrate on ensuring that the work of TCALL reflects and is reflected in the appropriate standards and trends in technology.
Orange.
ORANGE essentially provides a support role and is undertaking specific study tasks.
A brief summary of the support provided during 1997 is outlined, together with a note on some research into video transmission technology.
Dawn Bradford’s (CCR) development of the SMS software generated numerous technical questions relating to the air-interface protocols, handsets and data cards. Orange has provided information wherever possible.
Earlier this year, TCALL encountered problems with fax and data calls which was investigated and found to be due to the laptops that were being used to initiate the data calls. After much investigation it was found that there was a compatibility problem.
Orange has provided billing information required for the TCALL user trials carried out by CDS.
For Portset Systems Ltd., Orange has undertaken to look at the possibility of using a laptop to initiate normal voice related calls.
Orange has carried out video package studies as part of their contribution to the project, Using the GSM Network for Video Telephony. It was found that two of the three GSM initiatives (High Speed Circuit Switched Data and the Enhanced 14.4 kbits/sec Channel Coding) plus the ORANGE videophone have the potential to allow video telephony of sign over a GSM network. However, costs associated with the telecommunication of video (about 3 times regular use) may limit the use made of the service by the deaf.
Additionally, Orange has provided ad-hoc support for the TCALL project team.
Portset Systems Ltd.
Although two separate items were conceived in the early days of the project, the development of the TTS and STT telephone switchboard has been amalgamated into one unit. A draft proposal for a central server / switchboard device to process both TTS and STT telephone traffic was presented at the end of 1996 and a working system developed during 1997.
Text-to-Speech synthesis technology is well advanced and the core engine can be customised through selection of the voice parameter settings to suit the nationality, sex and accent required of the speaker.
At 1997, Speech-to-Text recognition is in a comparative infancy and extensive computer resources are needed to work with the multitudinous ways that people have of speaking. The TCALL requirement is for an ability to handle large vocabularies and speaker independence, something not yet possible.
The prototype switchboard TTS / STT was presented to the November meeting. Although in a rudimentary form due to limitations of the ASR engine, it attracted stimulating discussion. Plans are now in hand to modify the equipment, extending the recognition vocabulary and STT speed by reducing it to speaker-dependent capabilities and microphone (rather than telephone) input.
It had been assumed that suitable telephone management equipment would be available at reasonable cost for our application. But this proved not to be the case and Portset has, unexpectedly, had to undertake design and construction, from first principles, of its own interface processor. The telephone interface still remains a major problem for which no successful solution has yet been found, and requires extra funding to meet the expectations of our Partners.
In the Autumn of 1997, ORANGE provided Portset with an enhancer for connection with their cellular phone network. This, subsequently enabled Portset to evaluate TTS and STT testing using the Orange cellular network. We were pleased with the results obtained, not finding any loss of performance for telephone speech sent for ASR processing nor in transmission qualities of synthesised speech.
Achievements.
WorkPackage deliverables for the second year have been presented on time and represent both a collection of knowledge and research, and accomplishment of the second year’s project tasks. There were a number of ad-hoc reports introduced to TCALL: these represented work done outside the project but whose findings were relevant to the project. The enthusiasm of the three academic partners was recognised when each was awarded additional funding by the DTI for research work to provide addition knowledge to the project.
Problems.
Several problems arose during the year but it is felt all have been resolved, or soon will be. Among these could be reported: (a) Portset had been unable to integrate with the ORANGE network, a problem swiftly ended when ORANGE took the initiative of providing PSL with a signal enhancer. (b) ORANGE were not always happy with the security of use of their phones and airtime during trials. This was exacerbated by the increasing vigilance of HM Inland Revenue suspecting mobile phones to be a business ‘perk’. At the end of 1997, all parties are conscious of their respective responsibilities and the trials envisaged for 1998 will be subject to close control. (c) There have been comments that telephone air-time provision, both for the testing of project developments and for generally making the deaf community aware of mobile phone facilities, has been under-estimated. There have been fears that the project might go over-budget but caution is being exercised together with regular discussions with ORANGE as the service provider. (d) Portset found themselves committing more resources to the development of the TTS/ASR switchboard during 1997 than had originally been estimated. The telephone integration of the speech devices became a real challenge, as was the need to operate two very distinct sound cards within one PC environment.
Year 3 intentions.
The principle targets for the third year of the project in respect of each Partner are:-
Bristol - CCR: further development of TextTalk, with the possibility of a commercially available version, plus continued technical support to CDS during user trials.
Bristol - CDS: will be continuing to undertake extensive user trials.
Dundee - CME: continuing assistance to Partners with technical questions, experimental design and evaluation procedures. CME has also undertaken to ensure that the work of TCALL reflects, and is reflected in, the appropriate standards and trends in technology.
ORANGE: continued project support by technical advice, telecommunications equipment and air-time.
Portset Systems Ltd: reconfiguration of the prototype TTS/STT switchboard-server to meet the requirements of CDS for evaluation in trials within the deaf community.
This is the final year of the project and each Partner is giving consideration to the presentation of individual and collective results from the research and development work carried out.
Conclusions.
The Partners continue to co-operate well together. There is a good working relationship which enables areas of conflict to be discussed openly and resolved professionally.
The Partners bring experience of their other interests to the project and this extensive cross-fertilisation stimulates and inspires the group. CDS make sure the Partners focus on, and understand the benefits that TCALL work generates for the deaf community, even as opportunities to extend our knowledge to other more commercial applications become apparent.
The project work packages are to schedule and there is considerable enthusiasm to see our experience realised for the benefit of the deaf community.
The project has consumed more resources than was budgeted by some of the Partners.
The Partners have been instructed to consider ways of disseminating the results and enthusiasm of the project. There is also the need to make a wider public aware of the value telecommunication have for the deaf and for speech disabled people. These questions will be addressed during the coming, the final year of the project.
Graham Thomas,
TCALL Project Manager.
January 1998