Problem to be solved: to provide a speech enhancement apparatus, a speech enhancement method and a speech enhancement program for improving recognition accuracy of voice data.Speech enhancement deviceCalculated from body vibration of speakerThe pitch component during the utterance of the speakerThe generator generates a multimodal data based on the frequency spectrum of the sound data obtained by detecting the utterance of the speaker by an air conduction microphone.To the learning result obtained by learning the correspondence between pregenerated multimodal data and frequency spectrum of voice data.By entering the newly generated multi modal dataAn output section for outputting a frequency spectrum of voice data is provided.Diagram
展开▼