Improving Speech Recognition for the Elderly: A New Corpus of Elderly Japanese Speech and Investigation of Acoustic Modeling for Speech Recognition

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 6578–6585 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 6578 Improving Speech Recognition for the Elderly: A New Corpus of Elderly Japanese Speech and Investigation of Acoustic Modeling for Speech Recognition Meiko Fukuda1, Hiromitsu Nishizaki2, Yurie Iribe3, Ryota Nishimura1, Norihide Kitaoka4 1Tokushima University, Tokushima, Japan 2University of Yamanashi, Kofu, Japan 3Aichi Prefectural University, Nagoya, Japan 4Toyohashi University of Technology, Toyohashi, Japan fukuda.meiko@tokushima-u.ac.jp, hnishi@yamanashi.ac.jp, iribe@ist.aichi-pu.ac.jp, nishimura@is.tokushima-u.ac.jp, kitaoka@tut.jp Abstract In an aging society like Japan, a highly accurate speech recognition system is needed for use in electronic devices for the elderly, but this level of accuracy cannot be obtained using conventional speech recognition systems due to the unique features of the speech of elderly people. S-JNAS, a corpus of elderly Japanese speech, is widely used for acoustic modeling in Japan, but the average age of its speakers is 67.6 years old. Since average life expectancy in Japan is now 84.2 years, we are constructing a new speech corpus, which currently consists of the utterances of 221 speakers with an average age of 79.2, collected from four regions of Japan. In addition, we expand on our previous study (Fukuda, 2019) by further investigating the construction of acoustic models suitable for elderly speech. We create new acoustic models and train them using a combination of existing Japanese speech corpora (JNAS, S-JNAS, CSJ), with and without our ‘super-elderly’ speech data, and conduct speech recognition experiments. Our new acoustic models achieve word error rates (WER) as low as 13.38%, exceeding the results of our previous study in which we used the CSJ acoustic model adapted for elderly speech (17.4% WER). Keywords: Japanese speech corpus, elderly, acoustic modeling 1. Introduction In recent years in Japan, the use of the internet has increased among the elderly. A survey conducted by the Ministry of Internal Affairs and Communications found that the percentage of elderly people who searched for information on the internet within the past year was 76.6% for those aged 60-69, 51.0% for those aged 70-79, and 21.5% for those aged 80 and older (Ministry of Internal Affairs and Communications, 2019). However, as their visual acuity and hand motor function declines, it becomes increasingly difficult for the elderly to operate PCs and smartphones, which could result in a digital divide between the elderly and the younger generation. As governmental and corporate services are increasingly provided via the internet, this access problem will become more severe. We believe that speech recognition technology is a promising modality for addressing this problem, allowing the elderly to maintain access to information through their electronic devices. Furthermore, as society ages labor shortages are becoming apparent in Japan. As one countermeasure, a joint project between the Ministry of Health, Labor and Welfare and the Ministry of Economy, Trade and Industry is underway to promote the development and deployment of robots for assisting the elderly with their daily activities. In 2017, assisted living devices using robot technology …

Content truncated to 500 words