Browsing by Author "Paajanen, Auri"

Aims. In face-to-face communication, information contained in speech is also transmitted visually via observing articulation movements on the speaker’s lips and face. Silent lipreading is difficult since many of the phonemes of spoken language look visually similar and thus fall within the same viseme categories. Lipreading at single word level requires at least some of the phonemes to be recognized with sufficient accuracy in order for lexical activation to reach the correct target word in the observer’s mental lexicon. Word-level lipreading is also guided by language specific phonotactics, but its’ contribution to lipreading is not well understood. Lipreading ability shows great individual variation in both normal hearing and hearing-impaired populations. There is currently a lack of lipreading assessment tools in logopedic clinical practice in Finland. The aim of this study was to develop a Finnish single word lipreading test that can be delivered via a web application. I examined whether word videos produced in the Huulioluku research project can be used to measure lipreading ability in normal hearing adults and which test procedure appears to be functional. In addition, viseme analysis was used to examine features that contributed to the lipreadability of individual test words. Methods. The lipreading task used in this study consisted of word videos (43 Finnish words containing 2-3 syllables) developed in the Huulioluku research project. The participants (N=60) were normal hearing adults aged from 18 to 50. The data was collected remotely in March 2021 via a web application. Participants viewed silent word videos spoken by a female speaker and indicated which word had been said via a free text response. A point was given to each response that matched the target entirely. Deviation in participants’ scores (max 43 points) and the proportion correct for individual words were analysed. In a qualitative part of the study, the most frequent erroneous responses for each individual word were collected and a viseme comparison between the target and error response was carried out. Results and conclusions. The lipreading test showed great individual variation in the participants’ lipreading ability (score range between 1 and 34). The percentage correct for individual words ranged from 0 to 78 %. A short test of 20 best recognised words also showed great individual variation, with scores ranging from 20 to 1, suggesting that it could be used as such in future test developments. As seen in previous research, words with the least amount of visual competitors in the lexicon were recognized with the most accuracy. Consonants /v/ /p/ and /s/ were highly recognizable, whereas /r/ and /n/ were poorly recognized. The ability to perceive phoneme duration accurately as short/long proved important in recognition of Finnish words. Vowel discrimination was facilitated by a feature in the Finnish language phonotactics – vowel harmony. The words used in the test assessed the ability to lipread Finnish language specific phonological, morphological and prosodic features, which suggests that the 20-word test may be a valid tool.