ACOUSTIC MODEL CONDITIONING FOR SOUND CHARACTERISTIC

ACOUSTIC MODEL CONDITIONING FOR SOUND CHARACTERISTIC

To provide a speech processing system and a method improving inaccuracy in speech recognition capability due to a rare environmental condition.SOLUTION: A method captures a segment of speech voice having a key phrase voice and speech production continuing immediately afterward, has an encoder calcul...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: ZIZU GOWAYYED, KEYVAN MOHAJER
Format: Patent
Sprache:eng ; jpn
Schlagworte:
ACOUSTICS
MUSICAL INSTRUMENTS
PHYSICS
SPEECH ANALYSIS OR SYNTHESIS
SPEECH OR AUDIO CODING OR DECODING
SPEECH OR VOICE PROCESSING
SPEECH RECOGNITION
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:To provide a speech processing system and a method improving inaccuracy in speech recognition capability due to a rare environmental condition.SOLUTION: A method captures a segment of speech voice having a key phrase voice and speech production continuing immediately afterward, has an encoder calculate sound embedding by using a segment corresponding to the key phrase, and estimates phenom from an utterance sound signal by using a model which is a sound model for voice recognition conditioned to sound embedding as input.SELECTED DRAWING: Figure 4A 【課題】珍しい環境条件による、スピーチ認識能力の不正確さを改善したスピーチ処理システム及び方法を提供する。【解決手段】方法は、キーフレーズ音声と、そのすぐ後に続く発話と、を有するスピーチ音声のセグメントをキャプチャし、エンコーダが、キーフレーズに対応するセグメントを用いてサウンド埋め込みを計算し、音声認識のための音響モデルが、入力としてのサウンド埋め込みに対して条件付けされたモデルを用いて、発話音声信号からの音素を推定する。【選択図】図４Ａ