Speech recognition data enhancement method based on feature clipping and translation

The invention relates to a speech recognition data enhancement method based on feature clipping and translation, and belongs to the technical field of speech recognition processing. According to the method, audio signal features are cut and translated in a time dimension and a frequency dimension re...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: CHENG HAODONG, GUO YUHANG, CHEN SHUOYING, WU LITING
Format: Patent
Sprache:chi ; eng
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:The invention relates to a speech recognition data enhancement method based on feature clipping and translation, and belongs to the technical field of speech recognition processing. According to the method, audio signal features are cut and translated in a time dimension and a frequency dimension respectively. Wherein in the time dimension, the characteristic value of a certain time period is randomly selected, the characteristic of the time period is cut, and then the characteristic value which is not cut is translated along the time dimension direction; in the frequency dimension direction, the characteristic value of a certain frequency band is randomly selected, then the characteristic of the frequency band is cut, and the characteristic value which is not cut is translated along the frequency direction. According to the method, an audio signal does not need to be regenerated, features do not need to be extracted, waste of storage space and operation time is avoided, the data enhancement effect is better