Audio comparison system and audio energy comparison method thereof

Abstract

The invention discloses an audio comparison system and an audio energy comparison method thereof. The audio comparison system comprises an audio sequence determining unit, an audio sequence processing unit and an audio similarity judging unit. The audio energy comparison method comprises the steps of removing an abnormal frame to acquire an energy mean ratio (K value) and then counting points which are more than the K value in a normal interval to judge the similarity of two paths of audio frequencies. The audio comparison system of the invention integrally partitions the audio frequencies to be compared into a plurality of audio frames and uses the energy comparison among the audio frames to acquire the similarity of the compared audio frequencies, so the comparison mode is little influenced by the outside environment, does not cause error report phenomenon even under the environment with large noise and large disturbance, and has significance for audio transmission on broadcast.

Description

technical field [0001] The invention relates to the field of voice recognition, in particular to an audio recognition system and an audio comparison method thereof. Background technique [0002] Traditional speech recognition is to recognize a specific content from speech and generate other related applications according to the content. Existing audio comparison technologies are based on the Mel Frequency Cepstrum Coefficient (MFCC) algorithm, which extracts audio feature parameters through the MFCC algorithm, and then determines audio similarity based on the Euclidean distance of the audio feature parameters. In use, it has been shown that such a method can work normally when the audio quality is high, but there are certain false alarms in the occasions with high noise and interference, especially when the audio content is music. [0003] In practical applications, the application of comparison is usually to compare the same program obtained through different signal receiv...

AI Technical Summary

Problems solved by technology

Existing audio comparison technologies are based on the Mel Frequency Cepstrum Coefficient (MFCC) algorithm, which extracts audio feature parameters through the MFCC algorithm, and then determines audio similarity based on the Euclidean distance of the audio feature parameters. In use, it has been shown that such a method can work normally when the audio quality is high, but there are certain false alarms in the occasions with high noise and interference, especially when the audio content is music.

[0003] In practical applications, the application of comparison is usually to compare the same program obtained through different signal receiving methods, such as the comparison between satellite programs and air-received programs, and different signal receiving methods will inevitably lead to two channels of audio content There are certain differences. Due to the interference of external factors, the received content is also different from the source content. The MFCC algorithm extracts audio feature parameters according to the auditory characteristics of the human ear, so the difference in audio content will cause the algorithm to generate false positives.

Images