Automatic time correcting method for recording file

Abstract

The invention provides an automatic time correcting method for a recording file. The method comprises a step of reading LKJ alarm data through standard tone, and generating a model database, a step of removing noise in the recording file, a step of extracting all voice clips in the recording file and the times corresponding to the voice clips, a step of extracting a voice clip whose interval time between adjacent voice clips is in a set interval threshold range and whose time lengths is in a set time length threshold range as an effective voice clip, a step of matching the effective voice clip and the data in the model database, obtaining alarm information in the effective voice clip, and obtaining the alarm category of the alarm information, and a step of comparing the times to which the obtained alarm information and the alarm information of the LKJ belong, obtaining the delay of the recording file, and carrying out time correction on the recording file according to the delay.

Description

technical field [0001] The invention relates to a time checking method, in particular to a time checking method for recording files. Background technique [0002] Since the recording file is an audio signal recorded on the spot when the locomotive is running, there are nonlinear and complex noise signals such as whistle, flight attendant's reply, locomotive mechanical vibration, etc., which greatly increases the difficulty of speech recognition. It is difficult to accurately identify LKJ voice alarm information by using traditional Mel cepstrum frequency (MFCC) to extract signal frequency features and hidden Markov (HMM) algorithm for identification. Due to the problem of the internal crystal oscillator of the recording pen, there is often a deviation between the time of the recording file and the reference time axis of LKJ, so it is impossible to perform audio and video linkage analysis. Contents of the invention [0003] The technical problem to be solved by the present...

AI Technical Summary

Problems solved by technology

[0002] Since the recording file is an audio signal recorded on-site when the locomotive is running, there are nonlinear and complex noise signals such as whistle, flight attendant's reply, locomotive mechanical vibration, etc., which greatly increases the difficulty of speech recognition

Using traditional Mel cepstrum frequency (MFCC) to extract signal frequency features and hidden Markov (HMM) algorithm for recognition is difficult to accurately identify LKJ voice alarm information

Due to the problem of the internal crystal oscillator of the recording pen, there is often a deviation between the time of the recording file and the reference time axis of LKJ, and it is impossible to perform audio and video linkage analysis

Images