音声ブラウザご使用の方向け: SKIP NAVI GOTO NAVI

Web Posted on: August 4, 1998


Evaluation of the OSCAR hand held two-channel vibro-tactile speech communication aid for the deaf

Karl-Erik Spens
Eva Agelfors
Ulla Lignell*

Dept. of Speech, Music and Hearing
KTH, SE-100 44 Stockholm
tel: +46 8 790 7593
fax: +46 8 790 7854
email: karl-erik.spens@speech.kth.se
*Dept. of Audiology, Sahlgrenska Hospital, Gothenburg

 

1. Summary

Within the OSCAR-project (Optimal Speech Communication Assistance for Residual Abilities) within the TIDE frame a very flexible speech processing hearing aid has been developed for the deaf and for people with profound hearing losses. The processed signals displayed to the user could be auditive and or tactile. This paper concerns only the tactile tactile part of the project.
The main idea of the aid is that it should not be a general tactile "hearing aid", but rather give optimum tactile support to lip-reading in a close communication situation for those who are adventitiously deaf and have a rather good knowledge of the language structure. It should be used when speech communication is necessary and other methods of communication i.e. sign language, lip-reading, reading and writing etc. fail.
The aid conveys two types of information via its two vibro-tactile transducers, the syllabic rhythm and friction. The main result is that the aid does convey significant complementary information to lip-reading. However, the acceptance of the aid is varying and depending of the users personal characteristics.



| Top |

2. Introduction

Most tactile aids available on market today (MiniVib4, TAM, Tactaid II and VII) are using transducers supposed to be attached to the body, the wrist, sternum, the clavicle, the arm etc. If the aid serves the purpose of being prosthesis for "hearing" used on an every day basis with a minimum of cosmetic load, this is probably the most appropriate trade off between a global monitoring of the acoustic environment and a good support for lip-reading. In some cases the inconvenience of using the hand and it's higher spatial and temporal resolution is worth while and then the optimal design will be different. The OSCAR tactile aid is a hand held device, which should be used only in those situations when speech communication is preferred and other modes of communication fail. It is one hand held torch like self-contained unit worn in the pocket and used when needed. The microphone can be directional and it can be held in a very

Fig 1. Schematic and real picture of the hand held tactile OSCAR speech communication aid. The aid is held like a torch. The microphone is directed towards the speaker and the hand and thumb is stimulated tactually by the two respective transducers.

natural way close to the talker's mouth, thereby profiting from a better signal-to-noise ratio. Other signal sources like TV, the SIVO aid (Faulkner et al 1996) and lip-reading training equipment, are an option. There is sufficient space for battery capacity and elaborate circuitry. The two tactile transducers are designed to stimulate different receptors, thereby avoiding masking effects between them Spens et al (1997) (fig. 1)

 

2.1. Displayed tactile information

Prosodic information is an important factor for the intelligibility of speech in general and Öster (1990) has shown that the most important factor for the intelligibility of the speech of hard of hearing and deaf children is prosody. For a lip-reader only the very poor prosodic information displayed visually by the talker's lips is available. The Swedish tactile communication aid MiniVib4 conveys the syllabic rhythm via a single transducer. To evaluate such an aid, stimuli must contain a large portion of prosodic information i.e. it must be sentence material. Segmental stimuli like VCV and monosyllabic words would show almost no gain with this type of aid. Another implication is that such an aid will not work well in a speech-training situation because the aid's ability to discriminate between phonemes is limited. Speech Tracking or Connected Discourse Tracking (CDT) (De Filippo and Scott 1978) is an evaluation method used for connected speech. An investigation in Gothenburg (Axelsson et al 1986) with six deaf subjects showed an increased average tracking rate from 20 to 28 words per minute when using the MiniVib tactile aid (Spens 1995).

The tactile OSCAR aid has two channels, partly to match the number of features displayed by the audio processor (SiVo-3) also developed within the OSCAR-project (Faulkner 1996). The main speech feature conveyed to the hand is the syllabic rhythm for reasons given above. The second channel conveys fricative information for reasons of reliable extraction and high information content.

 



| Top |

3. Results

The Oscar aid was tested in a connected speech condition using a modified speech tracking procedure (Gnosspelius & Spens 1992). All results are preliminary in the respect that the amount of training was short. It is shown by (Spens 1995) that the average error free communication rate (=maximum communication rate) will directly influence the speech tracking results in a way that can not be easily predicted. Therefore all tracking results were measured at a constant error free rates. This rate can otherwise vary between sessions at random, at will or because of bias. It is therefore essential to use the same error free communication rate when comparing tracking results.

Fig. 2 shows the results of unaided and aided tracking under different conditions for a normal hearing subject with a simulated deafness. The left panel training time was (12+24+24+12)*5 min. and the right panel was (6+20+20+6)*5 min., in total about 10 hours. The unaided post-tests were added to ensure that learning did not play a major role. The 2-channel OSCAR aid exhibits a superior performance compared to a single-channel aid (MiniVib4) as well as a two-channel aid TactAid-II with two transducers with the same mode of stimulation.


Fig. 3 shows results from four Swedish subjects with profound hearing losses. They have all improved their tracking skill when using the Oscar (TA) tactile aid. However, they performed very differently. The unaided condition was compared to the hearing aid condition, one or more tactile conditions and sometimes also the combined tactile hearing aid condition. Subjects IB and MM are very good lip-readers. They performed very well both in the unaided and in the aided conditions and obtained tracking scores close to the maximum error free rate (70 words/min). IB improved her tracking rate by about 3w/m or 6% compared to the unaided condition. She also tried the version controlled by the SIVO (Faulkner et al 1996) and achieved about the same result. She had almost no benefit from her hearing aid. Subj. MM had more benefit from her hearing aid than from the tactile aid. However, she claimed she "heard" the sounds stronger when she used the tactile aid in combination with her hearing aid. The tactile aid added 2 words/min. to her performance compared to both the unaided condition and when she used her hearing aid. IT and MM both had the opinion that tactile support did not add enough to their performance. Subj. AK is an average lip-reader with almost no benefit from her hearing aid and therefore a lower error free rate (60 words/min) was used. She improved her tracking rate from 24 w/m in the unaided condition to 33 w/m with the tactile aid. A single channel tactile aid, MiniVib4 gave her about half the benefit compared to the two-channel OSCAR aid. Her subjective response to the tactile aid was very positive.

Subject UW is an average lip-reader who never uses a hearing aid. She has on a daily basis been using a single-channel tactile aid (MiniVib4) for 12 years and gets good benefit from this. However, she claims that she is so used to her MiniVib that the new OSCAR aid is somewhat disturbing. Interestingly she gets somewhat better results with the OSCAR aid after only one hour of use. Her training will continue.

Also use under non-speech conditions are reported. One English subject gets the most important help from the detection of noise sources in the home, and one Dutch subject is very enthusiastic about his new ability to perceive high frequency sounds of violins.



| Top|

4. Discussion

All subjects indicate improvements in tracking performance when using the OSCAR aid compared to the unaided condition. (See figs. 2 and 3). For those subjects who have tested other commercially available tactile aids like MiniVib4 and Tactaid II, the OSCAR aid produced better results. Retrospectively it can be anticipated that the two very good Swedish lip-readers would show a little more improvement in case they were given a more difficult test material. Their tracking performance in the test was too close to the error free rate, and hence probably suffering from ceiling effects. A more difficult text material would also most probably increase their motivation to attend more to the tactile stimulation.

Test results from connected speech conditions (figs 2 and 3), and spontaneous comments from the subjects about much "clearer" tactile stimulation support the hypothesis of less masking between different types of transducers (Spens et al. 1996).

It is the author's impression that an increased number of transducers do not correspond to an increased benefit of a multi-channel tactile communication aid used in a connected speech situation (Spens 95). If the aid is used as a speech training aid for slowly presented syllables, single words or very slow (clear speech), that is another purpose of the aid, and then a multi-channel aid could perform well. The main reason for the unsuccessful increase of the number of channels (transducers) in tactile aids used for connected speech, is believed to be sensory masking between the tactile channels. Huss & Spens (1996) and Carney & Beachler (1986) support this. All OSCAR results reported in this paper are obtained after short periods of training. Still they indicate benefit from a tactile aid and better performance when using two similar transducers and even better if two different transducers are used. Learning effects are also indicated and it could be expected that benefit would slowly improve with long-term training. However, data about this is something we would like to have, but still lack.



| Top|

Literature

Axelsson, A., Berenstaf, E. och Spens, K-E.: (1986.). "Erfarenheter av träning med ett vibrotaktilt hjälpmedel." Kursbok för Nordiska Audiologiska Sällskapets möte i Åbo, 167-169,.

Carney, A. E., Beachler, C. R. (1986). "Vibrotactile perception of supra segmental features of speech: A comparison of single-channel and multi-channel instruments." J. Acoust. Soc. Am. 79(1): 131-140.

De Filippo, C.L. and Scott. B.L. (1978) "A method for training the reception of ongoing speech", J. Acoust. Soc. Am. 63, pp. 1186-1192.

Faulkner, A. (1996): "The TIDE project OSCAR" ISAC, Sint-Michielsgestel, The Netherlands. 4th International Sensory Aid Conference, Saint-Michielsgestel, The Netherlands,." Scandinavian Audiology, Vol. Suppl 47, 38-44

Gnosspelius, J. & Spens, K-E. (1992): "A computer based Speech Tracking procedure," STL-QPSR No. 1, pp. 131-137 (KTH, Stockholm)

MiniVib4: Tactile aid from Special Instruments Development, Box 3050, S-181 03 LIDINGÖ, Sweden. Phone +46-8-7662430

Spens, K.E. (1995). Evaluation of speech tracking results: some numerical considerations and examples. In Plant, G. & Spens, K.E. (eds.) Profound deafness and speech communication. London: Whurr Publishers Ltd., ISBN 1-897635-45-1

Spens, K. E., Huss, C., Dahlqvist, M. and Agelfors, E. (1997). "A hand held two-channel vibro-tactile speech communication aid for the deaf: characteristics and results. 4th International Sensory Aid Conference, Saint-Michielsgestel, The Netherlands,." Scandinavian Audiology, Vol. Suppl. 47: pp. 5-13

Tactaid II and VII, Tactile aids from Audiological Engineering Corporation, 35 Medford Street, Somerville, MA 02143-9925, USA

TAM, Tactile aid from Summit, Birmingham, UK

Öster, A.-M. (1990). Functional hearing and intelligibility of the speech of hearing impaired and deaf children. Int Congress of Audiology, Tenerife, Spain.



| Top|