2026-03-14T16:11:59Z https://kyutech.repo.nii.ac.jp/oai

oai:kyutech.repo.nii.ac.jp:00004371 2025-12-04T06:47:16Z 15:20

Probabilistic Prediction in Multiclass Classification Derived for Flexible Text-Prompted Speaker Verification 40178124 黒木, 秀一 Kurogi, Shuichi クロギ, シュウイチ 黒木 Kurogi クロギ 秀一 Shuichi シュウイチ Sakashita, Shota Takeguchi, Satoshi Ueki, Takuya 50612165 Matsuo, Kazuya 松尾, 一矢 マツオ, カズヤ Matsuo 松尾 マツオ Kazuya 一矢 カズヤ Copyright (c) Springer International Publishing Switzerland 2015 probabilistic prediction text-prompted speaker verification Gibbsdistribution-based extended Bayesian inference loss functions in multiclass classification So far, we have presented amethod for text-promptedmultistep speaker verification using GEBI (Gibbs-distribution based extended Bayesian inference) for reducing single-step verification error, where we use thresholds for acceptance and rejection but the tuning is not so easy and affects the performance of verification. To solve the problem of thresholds, this paper presents a method of probabilistic prediction in multiclass classification for solving verification problem.We also present loss functions for evaluating the performance of probabilistic prediction. By means of numerical experiments using recorded real speech data, we examine the properties of the present method using GEBI and BI (Bayesian inverence) and show the effectiveness and the risk of probability loss in the present method. 22nd International Conference on Neural Information Processing, ICONIP 2015, November 9-12, 2015, Istanbul, Turkey Springer 2015-11-09 eng conference paper AM http://hdl.handle.net/10228/5585 https://kyutech.repo.nii.ac.jp/records/4371 https://doi.org/10.1007/978-3-319-26532-2_24 978-3-319-26531-5 978-3-319-26532-2 0302-9743 1611-3349 Lecture Notes in Computer Science 9489 216 225 https://kyutech.repo.nii.ac.jp/record/4371/files/iconip15_216.pdf application/pdf 212.2 kB 2016-03-04