Double-threshold limited place name speech endpoint detection method

Abstract

The invention discloses a double-threshold limited place name speech endpoint detection method. Starting from a first frame signal, the energy of each frame voice signal and the minimum energy threshold value and the maximum energy threshold value are judged, and The zero crossing rate and the zero crossing rate threshold are determined to determine how the next frame signal is detected; in the case of possible access to voice status, by increasing the variable, the voice signals that appear in the speech section before the light time are reserved. Thedouble door limited place name speech endpoint detection method combines the characteristics of names of speech signal isolated words, improves double threshold of the traditional method, guarantee the first part of the speech signal light and short duration is not be judged as noise, so as to avoid the loss of speech signal, therefore the accuracy of endpoint detection and the adaptability of field application environment are improved, and the requirement of environment is reduced.

Description

technical field [0001] The invention belongs to the field of voice endpoint detection, in particular to a double-threshold place name voice endpoint detection method. Background technique [0002] With the rapid economic development and the increasingly prominent trend of globalization, the modern logistics industry has achieved unprecedented development in developed countries, and has produced huge economic and social benefits. Logistics resources include transportation, warehousing, sorting, packaging, distribution, etc. These resources are scattered in many fields, including manufacturing, agriculture, and distribution. [0003] In the sorting process, sorting is basically carried out manually at this stage. Since the workers are in a noisy working environment for a long time, they will inevitably have a certain sense of fatigue in their minds and bodies, and the singleness and repetition of work tasks will also make Their working conditions are too relaxed, which will i...

AI Technical Summary

Problems solved by technology

[0005] In speech recognition technology, endpoint detection technology is an extremely important link in speech recognition, and its effect directly affects the final recognition result. The traditional endpoint detection method based on short-term energy and zero-crossing rate is an ideal environment. It can only be applied in the medium, and for the place name speech signal of isolated words, the accuracy of endpoint detection is relatively low

Images