Annual Report 1999 Table of Contents 1999 in summary Speech Communication and Technology Centre for Speech Technology Speech Signal Processing Music Acoustics Hearing Technology National and International Contacts Faculty and Staff Teaching Finances, Research Grants Seminars and Conferences Arranged by TMH List of Publications TMH-QPSR Abstracts Arne Leijon Professor of Hearing Technology The long-term goal of research in Hearing Technology is to understand how the physiological state of the human auditory system affects the ability to detect and analyse sound, how an impaired ability to detect and analyse sound affects the ability to understand and produce speech, how personal aids should be designed and individually fitted to compensate for hearing impairment. The group pursues several applied projects concerning the design, fitting, and evaluation of personal aids for the hearing impaired, as well as methods and systems for teaching and training of speech perception and speech production. Multi-modal Sensory Aids Together with researchers in the speech group we are investigating a new method to support audio-visual telephone communication for hearing-impaired persons without special equipment for transmitting video information. The aim of this work is to generate an artificial talking face which can be used as a lip-reading support by the hearing-impaired telephone user. The articulation movements of the artificial face are controlled by parameters automatically derived from the speech signal. The control parameters will inevitably contain some errors caused by the automatic recognition system, and the critical question was whether the artificial face would still contribute to improve speech perception. Speech recognition tests with hearing-impaired listeners have now demonstrated that the artificial face can in fact contribute significantly, even when controlled by an imperfect recognition system which can probably be further improved in future work. Evaluation of Cochlear Implants The work on comparing results from conventional hearing aids and cochlear implants has been continued. The test battery includes measurement of time resolution, auditory, visual and audiovisual perception of consonants, vowels, prosodic features and speech tracking, as well as subjective assessment. The aim of the study is to obtain data to support criteria for patient selection for cochlear implant operations, and also to evaluate the relative importance of various speech cues as speech-reading support. Signal Processing Algorithms for Hearing Instruments An industry-related project investigates methods to improve the algorithm by which a digital hearing instrument adapts its characteristics automatically to varying acoustic listening conditions. A general pattern-recognition approach has been used in this work. A project (LISCOM) within the TIDE-program of the EU aims at improving listening comfort by noise suppression in hearing instruments and telephones. Our role in this project has been mainly to evaluate two different algorithms in various noise backgrounds. Laboratory test results indicate that it is possible to substantially reduce the subjective loudness of noise, but the speech intelligibility does not improve correspondingly. The psycho-acoustic evaluation methods developed within this project will also be used in other research. Individual Fitting of Advanced Hearing Instruments Several methods for computer-aided fitting of advanced (digital) hearing aids are being evaluated in a joint project including laboratory and clinical studies in Stockholm and in Lund/Kristianstad. Preliminary results have shown that different prescriptive methods, proposed in the scientific literature, lead to widely different recommendations. A special method for hearing-aid fitting, based on individual tests of masked short-tone thresholds, has been developed and tested at the hearing clinic at Karolinska hospital in Stockholm. We have co-operated with this work in order to evaluate some of the effects of basic auditory-signal analysis on speech recognition. This fitting method is also evaluated in the clinical study in Lund/Kristianstad. We are also participating in a reference group which aims to standardise and improve the hearing-aid fitting methods used in Swedish hearing clinics. This work has resulted in a manuscript for a textbook on hearing-aid fitting, which is intended for use in the education of clinical audiologists. The book will appear in print in the year 2000. Teaching and Training Systems We participate in an EU project (SPECO), developing a new audio-visual pronunciation teaching and training method and software system for speech- and hearing-impaired children. The basic part is a general language-independent measuring system and database editor. The database editor makes it possible to construct modules for all participant languages; these are English, Swedish, Slovenian and Hungarian. Two support systems are under development for its construction in all languages, one of them being for teaching and training vowels and the other for fricatives and affricates.