ОЦЕНКА ЭФФЕКТИВНОСТИ АКУСТИЧЕСКИХ МОДЕЛЕЙ ДЛЯ СИСТЕМ РАСПОЗНАВАНИЯ РЕЧИ НА ОГРАНИЧЕННОМ МАТЕРИАЛЕ. / Холявин, Павел Андреевич; Кочаров, Д.А.
2019. 45-52.Research output: Contribution to conference › Abstract
}
TY - CONF
T1 - ОЦЕНКА ЭФФЕКТИВНОСТИ АКУСТИЧЕСКИХ МОДЕЛЕЙ ДЛЯ СИСТЕМ РАСПОЗНАВАНИЯ РЕЧИ НА ОГРАНИЧЕННОМ МАТЕРИАЛЕ.
AU - Холявин, Павел Андреевич
AU - Кочаров, Д.А.
PY - 2019
Y1 - 2019
N2 - The paper presents an analysis of acoustic models for speech recognition trained on limited data. Two types of training material were employed: phonetically balanced texts and arbitrarily chosen fragments of read speech, each recording being about 3 minutes long. Test fragments consisted of recordings of both spontaneous and read speech. It is found that the main factor influencing the performance of the model is the number of sounds in the training material.
AB - The paper presents an analysis of acoustic models for speech recognition trained on limited data. Two types of training material were employed: phonetically balanced texts and arbitrarily chosen fragments of read speech, each recording being about 3 minutes long. Test fragments consisted of recordings of both spontaneous and read speech. It is found that the main factor influencing the performance of the model is the number of sounds in the training material.
M3 - тезисы
SP - 45
EP - 52
ER -
ID: 78476350