作者: Luis Serrano García , Sneha Raman , Inma Hernáez Rioja , Eva Navas Cordón , Jon Sanchez
DOI: 10.1016/J.CSL.2020.101168
关键词: Digital signal processing 、 Esophageal speech 、 Alaryngeal speech 、 Speech Therapist 、 Database 、 Intelligibility (communication) 、 Parallel corpora 、 Speech processing 、 Computer science 、 Laryngectomee
摘要: Abstract A laryngectomee is a person whose larynx has been removed by surgery, usually due to laryngeal cancer. After most laryngectomees are able speak again, using techniques that learned with the help of speech therapist. This termed as alaryngeal speech, and esophageal (ES) one several production modes. considerable amount research dedicated study wide range aims such helping therapists evaluation diagnosis, improving its quality intelligibility digital signal processing techniques. We present you database Spanish ES voices, named AhoSLABI, which designed allow development new support technologies for this impairment. The primarily consists recordings 31 (27 males 4 females) pronouncing phonetically balanced sentences. Additionally, it includes parallel sentences 9 healthy speakers (6 3 facilitate tasks require small corpora, voice conversion or synthetic adaptation. Apart from sentences, sustained vowels set isolated words, can be valuable on analysis, diagnosis evaluation. paper describes main contents database, recording protocols procedure, well labeling process. acoustic characteristics speaking rate, durations recordings, phones silences, other compared those reduced voices. In addition, we describe an experiment improve performance ASR system speakers. resource will made available scientific community hope used life laryngectomees.