Impact of Age in ASR for the Elderly: Preliminary Experiments in European Portuguese

Evaluation of a Context-Aware Voice Interface for Ambient Assisted Living

“…The results of these studies show that an adaptation of the AM to each speaker makes the ASR performances closer to that with nonaged speakers; however, this implies that the ASR must be adapted to each speaker. These studies were done for English and French, but for another study for European Portuguese, which confirmed that the chronological age is a global explanatory factor [Pellegrini et al 2012]. However, this last study also emphasizes that many other effects can also be responsible for ASR performance degradation, such as decline in cognitive and perceptual abilities [Baeckman and Small 2001;Fozard and Gordont-Salant 2001].…”

Development of Automatic Speech Recognition Techniques for Elderly Home Support: Applications and Challenges

“…Automatic recognition of elderly speech was mainly studied for English by Vipperla et al [38] and for Portuguese by Pellegrini et al [26]. These two studies confirmed that the performances of standard recognizers decrease in the case of aged speakers.…”

Improving Speech Recognition through Automatic Selection of Age Group – Specific Acoustic Models

“…Furthermore, both children and elderly people are more likely to interact with computers using everyday language and their own commands, even when a specific syntax is required [5][6][7][8][9]. As compared with young to middle-aged adults, significantly higher word error rates (WERs) have been reported both for children [10][11][12] and for elderly speakers [10,13,14]. Improvements in ASR performance have been reported when using AMs adapted to children [10][11][12] and to the elderly [10,13,14], respectively, and more and more children's (e.g.…”

“…As compared with young to middle-aged adults, significantly higher word error rates (WERs) have been reported both for children [10][11][12] and for elderly speakers [10,13,14]. Improvements in ASR performance have been reported when using AMs adapted to children [10][11][12] and to the elderly [10,13,14], respectively, and more and more children's (e.g. [15][16][17]) and elderly speech corpora suitable for training AMs (e.g.…”

“…Using the default models in the case of children and the elderly is expected to deteriorate the ASR performance dramatically (cf. [10][11][12][13][14]). However, having to make a manual selection is rather cumbersome from the usability point of view.…”