Method and test system for batch testing speech identification and text synthesis

Abstract

The invention relates to a method and test system for batch testing speech identification and text synthesis. The method comprises steps that (1), the existing ASR and TTS technology jar package (so package) is integrated into the test system; (2), a to-be-tested function is selected; (3), a configured Config configuration file push in the PC is configured to test equipment according to the selected to-be-tested function; (4), the information in the configured file is checked; (5), function test is performed; and (6), a corresponding script is run to calculate the result, and the result is compared with the corresponding standard test result. The test system is characterized by comprising a control module, a storage module, an input module, a file sending module, a speech output module, afile receiving module, a test comparison module and a test result output module. The method and the test system are advantaged in that the whole process of speech identification and text synthesis from sending test data to result statistics is integrated, and work efficiency is greatly improved.

Description

technical field [0001] The invention belongs to the related fields of testing methods and systems, and in particular relates to a method and testing system for batch testing speech recognition and text synthesis. Background technique [0002] With the maturity of voice technology, there are more and more smart speakers, mobile phone voice assistants, smart homes, smart cars and other products controlled by voice. It is more convenient to control devices based on voice, and it has more advantages in experience. However, product testing for speech recognition has also become a difficult point. At present, for large-scale speech recognition testing, there is no complete testing method and testing tool for TTS text synthesis testing. Therefore, only speech recognition and text-to-speech can be separated. Open it for testing; for example, if you want to judge whether the pronunciation after text recognition is correct, you can only test the voice input and text recognition ASR fi...

AI Technical Summary

Problems solved by technology

However, product testing of speech recognition has also become a difficult point. At present, for the test of large-scale speech recognition, TTS text synthesis test does not have a complete test method and test tool. Therefore, speech recognition and text-to-speech can only be separated. Open it for testing; for example, if you want to judge whether the pronunciation after text recognition is correct, you can only test the voice input and text recognition ASR first, and then send the recognized text to TTS for testing; in addition, the TTS test is at this stage More people rely on human ears to listen, whether they read wrongly, whether there is a pause, whether they did not read directly, etc.; but listening with human ears is very time-consuming and manpower for manual judgment

Images