35. Elderly Adults Read Speech Corpus (EARS)

Data DOI


Producer, Project

Prof. Norihide Kitaoka, Toyohashi University of Technology

(ATR phonetically balanced sentences: ATR-Promotions, Inc., Newspaper article data: the Mainichi Newspaper Co., Ltd.)


This corpus consists of Japanese read speech uttered by elderly people over 70 years old. The average age of speakers in the S-JNAS corpus is 67.6 years old, while the average age of speakers in this corpus is 83.3 years old.

Recording was conducted in a quiet room of nursing homes. The read sentences were the same as S-JNAS, but the number of utterances was reduced compared to S-JNAS in consideration of the burden on the speaker. One set of ATR's 503 phonetically balanced sentences *1 (50 or 53 sentences) or one set of newspaper article sentences (10 sentences) per one speaker were recorded.

Initially, the data are provided which collected in four prefectures: Aichi, Tokushima, Mie, and Chiba. Additional data will be provided in the future.

*1 Please refer to the [Copyright of this list] section of the "ATR 503 sentences" page before using the text. (This corpus uses the text with permission.)


123 speakers (34 males and 89 females, 70-99 years old)

Speech file format

WAV format (16 kHz, 16 bit, Mono)

Distribution media



For research purpose only


No fee


All documents are written in Japanese.

Speech sample for test listening

Phonetically balanced sentence

救急車が 十分に 動けず 救助作業が 遅れている

Newspaper article sentence

まだ 正式 に 決まっ た わけ で は ない ので

Go to corpora list