Chinese online audio and video caption generation method

• Main Authors: CHEN KANGYANG, WANG YU, XUE JING
• Format: Patent
• Language: Chinese; English
• Published: 22.01.2019
• Subjects: ACOUSTICS; ELECTRIC COMMUNICATION TECHNIQUE; ELECTRICITY; MUSICAL INSTRUMENTS; PHYSICS; PICTORIAL COMMUNICATION, e.g. TELEVISION; SPEECH ANALYSIS OR SYNTHESIS; SPEECH OR AUDIO CODING OR DECODING; SPEECH OR VOICE PROCESSING; SPEECH RECOGNITION

The invention discloses a Chinese online audio and video caption generation method, which comprises the following steps: S1, an audio data extraction step, a server receives audio and video files, extracts audio data and converts the audio data into a standard format; S2, noise reduction step, processing noise reduction on that audio data to obtain an audio file; S3, a data cut step, cutting end points of that audio file to obtain an audio sample; S4, a fragment recognition step, further segmenting the obtained audio samples to obtain speech fragments, recognizing the speech fragments, and sorting out the recognition results of all the audio data; S5, a caption generation step, integrating and analyzing that text and the corresponding time axis to obtain a caption file, and matching the caption and audio data according to the generated caption file. The method of the invention can automatically complete speech recognition and caption generation of audio and video information, and effectively compensates for t