WO2019169914A1 - 语音测试方法及装置 - Google Patents

语音测试方法及装置 Download PDF

Info

Publication number
WO2019169914A1
WO2019169914A1 PCT/CN2018/118976 CN2018118976W WO2019169914A1 WO 2019169914 A1 WO2019169914 A1 WO 2019169914A1 CN 2018118976 W CN2018118976 W CN 2018118976W WO 2019169914 A1 WO2019169914 A1 WO 2019169914A1
Authority
WO
WIPO (PCT)
Prior art keywords
tested
voice
test
board
file
Prior art date
Application number
PCT/CN2018/118976
Other languages
English (en)
French (fr)
Inventor
陈杨顺
毛跃辉
郑威
Original Assignee
珠海格力电器股份有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 珠海格力电器股份有限公司 filed Critical 珠海格力电器股份有限公司
Publication of WO2019169914A1 publication Critical patent/WO2019169914A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L15/00—Speech recognition
    • G10L15/01—Assessment or evaluation of speech recognition systems

Definitions

  • the present application relates to the field of voice testing, and in particular to a voice testing method and apparatus.
  • the current voice performance test method requires 10 or more testers to perform long-term instructions, and performs performance tests on the wake-up rate and recognition rate of the voice board, which takes a long time and is easy to make the tester tired, resulting in test results. error.
  • the embodiment of the present application provides a voice testing method and apparatus, so as to at least solve the technical problem that a test caused by a long time command by a plurality of testers in the existing voice test method takes a long time.
  • a voice test method including: acquiring an audio test file; performing a test task by playing the audio test file on the voice board to be tested according to a state of the voice board to be tested. Generating a test result for indicating the performance of the speech board to be tested.
  • the audio test file includes a first audio file for the wake-up test, wherein, according to the state of the voice board to be tested, performing the test task by playing the audio test file on the voice board to be tested includes: The wake-up rate of the voice board to be tested is obtained by performing a wake-up test on the voice board to be tested by playing the first audio file, in a case where the state of the voice board to be tested is the state to be awake.
  • the method before performing the wake-up test on the to-be-tested voice board by playing the first audio file, the method further includes: acquiring a state of the voice board to be tested; and detecting a state of the voice board to be tested; If the state of the voice board to be tested is the to-be-awakened state, triggering the performing the over-playing of the first audio file to perform a wake-up test on the voice board to be tested; if the state of the voice board to be tested is wake-up a state in which the third audio file is played on the voice board to be tested, and the voice board to be tested is controlled to exit the awake state; if the state of the voice board to be tested is a source playing state, the pair is tested.
  • the voice board plays a fourth audio file, and the voice board to be tested is controlled to stop playing the source, and after the voice board to be tested stops playing the source, the third audio file is played by playing the voice board to be tested. Controlling the voice board to be tested to exit the awake state.
  • performing a wake-up test on the voice board to be tested by playing the first audio file, and obtaining an awakening rate of the voice board to be tested includes: playing the first voice board by using a high-fidelity speaker An awakening word in an audio file; acquiring a first text returned by the speech board to be tested; determining whether the first text is the wake-up word; if the first text is the wake-up word, determining that the wake-up is successful; The number of times to get the awakening rate.
  • the method further includes: if the first character is not the wake-up word, playing the wake-up word again; acquiring the second text returned by the voice board to be tested; if the second text is The wake-up words are used to determine the success of the wake-up.
  • the method further includes: if the second character is not the wake-up word, repeatedly playing a fifth audio for forcibly waking up the voice board to be tested; if the user successfully wakes up within a preset number of times The voice board to be tested continues the wake-up test; if the voice board to be tested is not woken up after the preset number of times, the wake-up test is ended.
  • the audio test file includes a second audio file for identifying a test and an annotation file corresponding to the second audio file, wherein, according to a state of the voice board to be tested, playing by the voice board to be tested
  • the performing the test task includes: performing a recognition test on the voice board to be tested according to the label file of the second audio file, where the state of the voice board to be tested is an awake state, The recognition rate of the speech board to be tested.
  • performing the identification test on the voice board to be tested according to the label file of the second audio file, and obtaining the recognition rate of the voice board to be tested includes: playing the voice board to be tested through a high-fidelity speaker The instruction word in the second audio file; obtaining the text content returned by the speech board to be tested; determining whether the text content is the same as the information indicated in the annotation file; if the text content and the annotation file The information indicated in the same is the same, and the identification is determined to be successful; the recognition rate is obtained according to the number of successful recognitions.
  • the method further includes: if the text content is different from the information indicated in the annotation file, playing the instruction word again; if the text content returned by the to-be-tested speech board returns the annotation The information indicated in the file is the same, and the identification is successful; if the text content returned by the speech board to be tested is different from the information indicated in the annotation file, it is determined that the recognition fails.
  • the annotation file includes at least one of the following: an audio file name, a voice content, and a format.
  • a voice testing apparatus including: an acquiring unit, configured to acquire an audio test file; and a testing unit, configured to pass the tested according to a state of the voice board to be tested
  • the voice board plays the audio test file, and performs a test task.
  • the generating unit is configured to generate a test result for indicating performance of the voice board to be tested.
  • a storage medium comprising a stored program, wherein the program executes the voice test method described above.
  • a computer terminal comprising: a memory for storing a program; a processor for running the program, wherein the program runs the voice test method described above .
  • the audio test file is obtained, and the test task is performed by playing the audio test file on the voice board to be tested according to the state of the voice board to be tested, and generating a voice board for indicating the voice board to be tested.
  • the method of performance test results through the automatic voice performance test and statistics according to the state of the voice board to be tested, achieves the purpose of automatic voice performance test and unmanned operation, thereby realizing the reduction of testers and speeding up the test speed.
  • the technical effect further solves the technical problem that the test of the existing voice test method requires a plurality of testers to perform long-term instructions and takes a long time.
  • FIG. 1 is a schematic flow chart of an optional voice test method according to an embodiment of the present application.
  • FIG. 2 is a schematic flow chart of another optional voice testing method according to an embodiment of the present application.
  • FIG. 3 is a schematic diagram of an optional software interface according to an embodiment of the present application.
  • FIG. 4 is a schematic flow chart of still another optional voice testing method according to an embodiment of the present application.
  • FIG. 5 is a schematic flowchart diagram of still another optional voice testing method according to an embodiment of the present application.
  • FIG. 6 is a schematic flowchart diagram of still another optional voice testing method according to an embodiment of the present application.
  • FIG. 7 is a schematic structural diagram of an optional voice testing apparatus according to an embodiment of the present application.
  • a method embodiment of a voice test method is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings may be performed in a computer system such as a set of computer executable instructions, and Although a logical order is shown in the flowcharts, in some cases the steps shown or described may be performed in a different order than the ones described herein.
  • FIG. 1 is a voice test method according to an embodiment of the present application. As shown in FIG. 1 , the method includes the following steps:
  • Step S102 obtaining an audio test file.
  • the voice test method of the embodiment needs to rely on the audio test file and the annotation file corresponding to the audio test file, and the audio test file includes a forced wake-up audio file, a forced stop of playing the audio file, a forced close of the voice audio file, and test identification.
  • the sample audio file required for waking up; the annotation file is information corresponding to the audio file of the voice recognition item to be tested (including the test audio file name, the play content of the test audio file, and the format information corresponding to the test audio file), In this way, the computer can obtain the information returned by the voice board through the serial port and compare the information contained in the label file.
  • Step S104 Perform an test task by playing an audio test file through the sound board to be tested according to the state of the voice board to be tested.
  • the audio test file includes a first audio file (ie, an awakening word audio file) for the wake-up test, wherein the audio test file is played through the sound board to be tested according to the state of the voice board to be tested,
  • the test task includes: performing a wake-up test on the voice board to be tested by playing the first audio file when the state of the voice board to be tested is in a state to be awake, and obtaining the wake-up rate of the voice board to be tested.
  • the method further includes: acquiring a state of the voice board to be tested; detecting a state of the voice board to be tested; and if the state of the voice board to be tested is a state to be awake If the state of the voice board to be tested is awake, the third audio file is played through the voice board to be tested, and the voice board to be tested is controlled to exit the awake state; The state of the voice board is the source playing state, and the fourth audio file is played through the voice board to be tested, the voice board to be tested is stopped to play the source, and after the voice board to be tested stops playing the source, the voice board is played.
  • the third audio file controls the voice board to be tested to exit the awake state.
  • the voice board before testing a voice command, the voice board must be in a state to be awake, in order to conform to the principle of actual user use.
  • Automated test software firstly, it is necessary to detect whether the voice board is in a state to be awake. If the state of the voice board is awake, the computer is forced to close the voice audio file (ie, the third audio file mentioned above) by playing the high-fidelity speaker on the voice board. To let it exit the awake state; if the voice board is in the source play state, let the computer forcibly stop playing the audio file (ie, the fourth audio file mentioned above) by playing the high-fidelity speaker on the voice board, and wait for the voice board to stop playing the message. After the state of the source, playback is again forced to close the voice audio file (ie, the third audio file described above) to exit the awake state.
  • the wake-up test is performed by playing the first audio file to test the voice board, and obtaining the wake-up rate of the voice board to be tested includes: playing the wake-up word in the first audio file by using the high-fidelity speaker to test the voice board; acquiring the voice to be tested The first text returned by the board; determining whether the first text is an awakening word; if the first text is an awakening word, determining that the wakeup is successful; and according to the number of successful wakeups, obtaining an awakening rate.
  • the voice test method of this embodiment further includes: if the first text is not an awakening word, playing the wake-up word again; acquiring the second text returned by the voice board to be tested; and if the second text is an wake-up word, determining that the wake-up is successful.
  • the voice test method of the embodiment further includes: if the second text is not an awakening word, repeatedly playing the fifth audio for forcibly awakening the voice board to be tested; if the voice board to be tested is successfully awake within a preset number of times , the wake-up test is continued; if the voice board to be tested is not woken up more than the preset number of times, the wake-up test is ended.
  • the awakening rate of the voice board is tested, and the voice board is in a state to be awake. Make sure that the voice board is in the state of being awakened. At this time, let the computer play different wake-up words and audio files (that is, the first audio file mentioned above) on the voice board through the high-fidelity speaker, and the voice board corresponding to the voice signal obtained by the voice board returns through the serial port.
  • the returned text content is an awakening word, and if so, the number of wake-up successes is increased by one; if not the wake-up word, or the wake-up word is incorrect, the current audio file is repeatedly played once; if successful, the number of successful wake-ups Add 1 if the wakeup is unsuccessful, then the number of wakeup failures is increased by one; when the playback is not awake, the audio file is forcibly awakened (ie, the fifth audio file mentioned above), the number of repetitions is 10, as long as 10 If the voice board is awakened in the second time, the task will continue; when more than 10 strong wake-ups are completed, the task ends, and the software prompts “The current test environment is incorrect, please check the environment”.
  • the audio test file includes a second audio file for identifying the test and an annotation file corresponding to the second audio file, wherein the audio is played through the soundboard to be tested according to the state of the voice board to be tested.
  • the test file includes: when the state of the voice board to be tested is the awake state, the identification test of the voice board to be tested is performed according to the second audio file labeling file, and the recognition rate of the voice board to be tested is obtained.
  • the recognition test of the voice board to be tested is performed according to the second audio file annotation file
  • the recognition rate of the voice board to be tested includes: playing the instruction word in the second audio file by using the high-fidelity speaker to test the voice board; The text content returned by the voice board; whether the text content is the same as the information indicated in the label file; if the text content is the same as the information indicated in the label file, the identification is successful; and the recognition rate is obtained according to the number of successful recognitions.
  • the method further includes: if the text content is different from the information indicated in the annotation file, playing the instruction word again; if the text content returned by the speech board to be tested is the same as the information indicated in the annotation file, determining that the recognition is successful; The text content returned by the voice board to be tested is different from the information indicated in the label file, and it is determined that the recognition fails.
  • the recognition rate is tested, and the voice board must be awake first. After the voice board is woken up, the recognition rate test can be performed. There are also two opportunities for the recognition rate.
  • the high-fidelity speaker plays different test word audio files (ie the second audio file mentioned above) on the voice board, and the text content returned by the voice board and the audio file corresponding to the annotation file are played through the serial port.
  • Content JSON (JavaScript Object Notation, a lightweight data exchange format) is consistent, consistently recorded as success, inconsistent is recorded as failure; when the first recognition is successful, the number of recognition successes is increased by 1, for the next cycle If the first time is unsuccessful, repeat the play once, when both times fail, the number of recognition failures is increased by 1, and then the next cycle is performed.
  • the annotation file includes at least one of the following: an audio file name, a voice content, and a format.
  • Step S106 generating a test result for indicating performance of the voice board to be tested.
  • the software converts the number of successful wake-ups of the test into a wake-up rate, converts the success number into the recognition rate, generates the final test result, and then outputs it to the Excel spreadsheet.
  • the wake-up rate and the recognition rate are synchronized.
  • Corresponding voice content and JSON information can be used.
  • the automatic voice performance test and statistics can be performed according to the state of the voice board to be tested, and the purpose of the voice performance test automation and the unmanned operation is achieved, thereby realizing the technical effect of reducing the tester and speeding up the test speed. Furthermore, the technical problem that the test caused by the long-term instruction of multiple testers by the existing voice test method takes a long time is solved. Realize voice performance test automation and unmanned operation; achieve standardized test, high accuracy of voice performance test, test phenomenon can be reproduced; shorten voice performance test time; eliminate interference from subjective factors, standardization of test process.
  • Step A start.
  • the voice test method of this embodiment uses a high-fidelity speaker to play an audio file to simulate a human voice (more realistic and human voice), and a high-fidelity speaker is placed on a stand of 1.5 meters in height to simulate a person's mouth position. After determining the position of the high-fidelity speaker, open the software for testing.
  • the software interface is shown in Figure 3.
  • Step B establish a task.
  • the software of this embodiment can be run under the Python 2.7 environment, and the test needs to rely on the test audio file, the annotation file corresponding to the test audio, and the test audio file includes the forced wake-up audio file, the forced stop of playing the audio file, and the forced shutdown.
  • the voice audio file and the sample audio file required for the test identification and wakeup; the annotation file is the information corresponding to the audio file of the voice recognition item to be tested (including the test audio file name, the test audio file play content, the test audio file Corresponding JSON information), so that the computer can obtain the information returned by the voice board through the serial port and compare the information contained in the label file.
  • the software is tested according to the contents of the annotation file. How many lines are cyclically tested, each time as a task.
  • step C it is guaranteed to be in a state to be awakened.
  • the voice board Before testing a voice command, the voice board must be in a state to be awake before it meets the principle of actual user usage. As shown in FIG. 4, the steps of ensuring that the state is to be awake include:
  • Step c1 start.
  • step c2 the state of the voice board is obtained.
  • step c3 it is in a state to be awakened.
  • step c4 If no, go to step c4;
  • step c4 the audio that exits the wake-up is played.
  • Step c5 ending.
  • Automated test software firstly, it is necessary to detect whether the voice board is in the state to be awake. If the state of the voice board is awake, let the computer forcibly turn off the voice audio file through the high-fidelity speaker to the voice board to let it exit the awake state; When the board is in the source playing state, the computer is forced to stop playing the audio file through the high-fidelity speaker playing on the voice board. After the voice board stops playing the source state, the player temporarily closes the voice audio file to make it quit the wake-up. status.
  • Step D wake up the test.
  • the steps of the wake-up test include:
  • Step d1 start.
  • step d2 the wake-up word audio is played.
  • step d3 the state of the voice board is obtained.
  • Step d4 whether it is awakened.
  • step d11 If yes, go to step d11.
  • step d5 the wake-up word audio is played again.
  • step d6 the state of the voice board is obtained.
  • Step d7 whether it is awakened.
  • step d8 If no, go to step d8;
  • step d11 If yes, go to step d11.
  • step d8 the number of wakeup failures is +1.
  • step d9 a strong wake-up word audio is played.
  • Step d10 whether it is awakened.
  • step d9 If no, go to step d9;
  • step d11 the number of successful wakeups is +1.
  • Step d12 the end.
  • the voice board After the C step, ensure that the voice board is in the state of being awakened. At this time, let the computer play different wake-up words and audio files on the voice board through the high-fidelity speaker, and then the voice board judges the text corresponding to the obtained audio signal through the serial port, and then judges Whether the returned text content is a wake-up word, if yes, the number of wake-up successes is increased by one; if not the wake-up word, or the wake-up word is incorrect, the current audio file is played again once; if successful, the number of wake-up successes is increased by one, if awake If the operation is unsuccessful, the number of wakeup failures is increased by one; when the playback is not awake, the audio file is forcibly awakened, and the number of repetitions is 10 times. If the voice board is woken up within 10 times, the task is continued; When the second strong wakeup, the task ends, the software prompts "The current test environment is wrong, please check the environment.”
  • Step E identify the test.
  • the speech board To test the recognition rate, the speech board must be awakened first, and the recognition rate test can be performed after the speech board is awakened. As shown in Figure 6, the steps of the identification test include:
  • Step e1 start.
  • step e2 the instruction word audio is played.
  • step e3 the speech board recognition result is obtained.
  • step e4 whether the recognition is successful.
  • step e5 If no, go to step e5;
  • step e9 If yes, go to step e9.
  • step e6 the instruction word audio is played again.
  • step e7 the speech board recognition result is obtained.
  • step e8 whether the recognition is successful.
  • step e9 If yes, go to step e9;
  • step e10 If no, go to step e10.
  • step e9 the number of successes is +1.
  • step e10 the number of failures is +1.
  • Step e11 the end.
  • the high-fidelity speaker plays different test word audio files on the voice board.
  • the text content returned by the voice board through the serial port is consistent with the playback content and JSON of the audio file corresponding to the annotation file. If the first recognition is successful, the number of successful recognitions is increased by one, and the next cycle is repeated. 1, then proceed to the next cycle.
  • Step F whether the task is completed.
  • Step G generate an excel report.
  • the software will convert the number of successful wake-ups of the test into the wake-up rate, the number of successful recognitions is converted into the recognition rate, and then output to the Excel table.
  • the wake-up rate and the recognition rate are synchronized.
  • the voice test method of the embodiment adopts a GUI (Graphical User Interface) interface, and the operation is simple and easy to use; the test audio is played by a computer, and the data returned by the voice board is received through the serial port for analysis, and compared with the annotation file. Calculate the speech recognition test result, and output the result to the Excel table, the whole process is automated; then realize the parallel test of wake-up rate and recognition rate, greatly shorten the test time and improve the test efficiency, compared with the traditional voice performance test, personnel demand The number is reduced by 90% and the test time is reduced by 50%.
  • GUI Graphic User Interface
  • the voice testing apparatus includes:
  • the obtaining unit 702 is configured to obtain an audio test file
  • the testing unit 704 is configured to perform a test task by playing the audio test file on the to-be-tested voice board according to a state of the voice board to be tested
  • a generating unit 706 is configured to: Generating test results for indicating the performance of the speech board to be tested.
  • the audio test file includes a first audio file for the wake-up test
  • the testing unit 704 is configured to perform the following steps: playing the audio by using the audio board to be tested according to the state of the voice board to be tested. Testing the file and performing the test task: in the case that the state of the voice board to be tested is in a state to be awake, the user is tested by performing a wake-up test on the voice board to be tested by playing the first audio file to obtain the voice board to be tested. Awakening rate.
  • the testing unit 704 is further configured to: acquire a state of the voice board to be tested; and detect a state of the voice board to be tested; and if the state of the voice board to be tested is the to-be-awake state, triggering execution Performing a wake-up test on the voice board to be tested by playing the first audio file; if the state of the voice board to be tested is an awake state, playing the third audio file on the voice board to be tested, and controlling the The voice board to be tested exits the awake state; if the state of the voice board to be tested is the source play state, the fourth audio file is played on the voice board to be tested, and the voice board to be tested is controlled to stop playing the source. And after the playing the to-be-tested voice board stops playing the source, the third audio file is played on the to-be-tested voice board, and the voice board to be tested is controlled to exit the awake state.
  • the testing unit 704 is configured to perform the following steps: performing a wake-up test on the to-be-tested speech board by playing the first audio file, and obtaining an awakening rate of the to-be-tested speech board: the high-fidelity speaker is used to The soundboard plays the wake-up word in the first audio file; acquires the first text returned by the voiceboard to be tested; determines whether the first text is the wake-up word; if the first text is the wake-up word, Determining that the wakeup is successful; according to the number of successful wakeups, the wakeup rate is obtained.
  • the testing unit 704 is further configured to: if the first character is not the wake-up word, play the wake-up word again; acquire the second text returned by the voice board to be tested; if the second text is The wake-up word determines that the wake-up is successful.
  • the testing unit 704 is further configured to: if the second character is not the wake-up word, repeatedly play the fifth audio for forcibly waking up the voice board to be tested; if the user wakes up successfully within a preset number of times The voice board is tested, and the wake-up test is continued; if the voice board to be tested is not awake for more than the preset number of times, the wake-up test is ended.
  • the audio test file includes a second audio file for identifying a test and an annotation file corresponding to the second audio file, where the testing unit 704 is configured to perform the following steps according to the state of the voice board to be tested. Playing the audio test file on the voice board to be tested, and performing a test task: in the case that the state of the voice board to be tested is an awake state, the label file according to the second audio file is to be tested. The voice board performs a recognition test to obtain a recognition rate of the voice board to be tested.
  • the testing unit 704 is configured to perform the following steps: performing a recognition test on the to-be-tested speech board according to the annotation file of the second audio file, to obtain a recognition rate of the to-be-tested speech board: by using a high-fidelity speaker pair
  • the speech board to be tested plays the instruction word in the second audio file; the text content returned by the speech board to be tested is acquired; and the text content is determined to be the same as the information indicated in the annotation file; The text content is the same as the information indicated in the annotation file, and the identification is determined to be successful; the recognition rate is obtained according to the number of successful recognitions.
  • the testing unit 704 is further configured to: if the text content is different from the information indicated in the annotation file, play the instruction word again; if the text board to be tested returns the text content and the The information indicated in the annotation file is the same, and the identification is successful; if the text content returned by the speech board to be tested is different from the information indicated in the annotation file, it is determined that the recognition fails.
  • the annotation file includes at least one of the following: an audio file name, a voice content, and a format.
  • a storage medium comprising a stored program, wherein the program executes the voice test method described above.
  • a computer terminal comprising: a memory for storing a program; a processor for running the program, wherein the program runs the voice test method described above.
  • the disclosed technical contents may be implemented in other manners.
  • the device embodiments described above are only schematic.
  • the division of the unit may be a logical function division.
  • there may be another division manner for example, multiple units or components may be combined or may be Integrate into another system, or some features can be ignored or not executed.
  • the mutual coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interface, unit or module, and may be electrical or otherwise.
  • the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, may be located in one place, or may be distributed to multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of the embodiment.
  • each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.
  • the above integrated unit can be implemented in the form of hardware or in the form of a software functional unit.
  • the integrated unit if implemented in the form of a software functional unit and sold or used as a standalone product, may be stored in a computer readable storage medium.
  • a computer readable storage medium A number of instructions are included to cause a computer device (which may be a personal computer, server or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present application.
  • the foregoing storage medium includes: a U disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a removable hard disk, a magnetic disk, or an optical disk, and the like. .
  • the solution provided by the embodiment of the present application can be applied to the field of voice testing.
  • an audio test file is obtained.
  • the audio test file is played by playing the voice board to be tested.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computational Linguistics (AREA)
  • Health & Medical Sciences (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Human Computer Interaction (AREA)
  • Physics & Mathematics (AREA)
  • Acoustics & Sound (AREA)
  • Multimedia (AREA)
  • Debugging And Monitoring (AREA)

Abstract

一种语音测试方法及装置，其中，该方法包括：获取音频测试文件(S102)；根据待测语音板的状态，通过对待测语音板播放音频测试文件，执行测试任务(S104)；生成用于指示待测语音板性能的测试结果(S106)。

Description

语音测试方法及装置 技术领域
本申请涉及语音测试领域，具体而言，涉及一种语音测试方法及装置。
背景技术
针对一些语音部分开发的需要，在完成预计功能之后要验证其语音的可靠性，需要进行性能测试。目前的语音性能测试方法，需要10个或10个以上的测试人员进行长时间的念指令，对语音板的唤醒率和识别率进行性能测试，耗时长，容易使测试人员疲倦，导致测试结果有误差。
针对上述的问题，目前尚未提出有效的解决方案。
发明内容
本申请实施例提供了一种语音测试方法及装置，以至少解决由于现有的语音测试方法需要多个测试人员进行长时间的念指令造成的测试耗时较长的技术问题。
根据本申请实施例的一个方面，提供了一种语音测试方法，包括：获取音频测试文件；根据待测语音板的状态，通过对所述待测语音板播放所述音频测试文件，执行测试任务；生成用于指示所述待测语音板性能的测试结果。
可选地，所述音频测试文件包括用于唤醒测试的第一音频文件，其中，根据待测语音板的状态，通过对所述待测语音板播放所述音频测试文件，执行测试任务包括：在所述待测语音板的状态为待唤醒状态的情况下，通过播放所述第一音频文件对所述待测语音板进行唤醒测试，得到所述待测语音板的唤醒率。
可选地，通过播放所述第一音频文件对所述待测语音板进行唤醒测试之前，所述方法还包括：获取所述待测语音板的状态；检测所述待测语音板的状态；若所述待测语音板的状态为所述待唤醒状态，触发执行所述过播放所述第一音频文件对所述待测语音板进行唤醒测试；若所述待测语音板的状态为唤醒状态，通过对所述待测语音板播放第三音频文件，控制所述待测语音板退出所述唤醒状态；若所述待测语音板的状态为信源播放状态，通过对所述待测语音板播放第四音频文件，控制所述待测语音板停止播放信源，并在所述待测语音板停止播放信源之后，通过对所述待测语音板播放 所述第三音频文件，控制所述待测语音板退出所述唤醒状态。
可选地，通过播放所述第一音频文件对所述待测语音板进行唤醒测试，得到所述待测语音板的唤醒率包括：通过高保真音箱对所述待测语音板播放所述第一音频文件中的唤醒词；获取所述待测语音板返回的第一文字；判断所述第一文字是否为所述唤醒词；若所述第一文字为所述唤醒词，确定唤醒成功；根据唤醒成功的次数，得到所述唤醒率。
可选地，所述方法还包括：若所述第一文字不为所述唤醒词，再次播放所述唤醒词；获取所述待测语音板返回的第二文字；若所述第二文字为所述唤醒词，确定唤醒成功。
可选地，所述方法还包括：若所述第二文字不为所述唤醒词，重复播放用于强制唤醒所述待测语音板的第五音频；若在预设次数内成功唤醒所述待测语音板，继续进行所述唤醒测试；若超过所述预设次数未唤醒所述待测语音板，则结束所述唤醒测试。
可选地，所述音频测试文件包括用于识别测试的第二音频文件及所述第二音频文件对应的标注文件，其中，根据待测语音板的状态，通过对所述待测语音板播放所述音频测试文件，执行测试任务包括：在所述待测语音板的状态为唤醒状态的情况下，根据所述第二音频文件所述标注文件对所述待测语音板进行识别测试，得到所述待测语音板的识别率。
可选地，根据所述第二音频文件所述标注文件对所述待测语音板进行识别测试，得到所述待测语音板的识别率包括：通过高保真音箱对所述待测语音板播放所述第二音频文件中的指令词；获取所述待测语音板返回的文字内容；判断所述文字内容与所述标注文件中指示的信息是否相同；若所述文字内容与所述标注文件中指示的信息相同，确定识别成功；根据识别成功的次数，得到所述识别率。
可选地，所述方法还包括：若所述文字内容与所述标注文件中指示的信息不相同，再次播放所述指令词；若所述待测语音板再次返回的文字内容与所述标注文件中指示的信息相同，确定识别成功；若所述待测语音板再次返回的文字内容与所述标注文件中指示的信息不相同，确定识别失败。
可选地，所述标注文件包括以下至少之一：音频文件名、语音内容、格式。
根据本申请实施例的另一方面，还提供了一种语音测试装置，包括：获取单元，用于获取音频测试文件；测试单元，用于根据待测语音板的状态，通过对所述待测语音板播放所述音频测试文件，执行测试任务；生成单元，用于生成用于指示所述待测语音板性能的测试结果。
根据本申请实施例的另一方面，还提供了一种存储介质，所述存储介质包括存储的程序，其中，所述程序执行上述的语音测试方法。
根据本申请实施例的另一方面，还提供了一种计算机终端，包括：存储器，用于存储程序；处理器，用于运行所述程序，其中，所述程序运行时执行上述的语音测试方法。
在本申请实施例中，采用获取音频测试文件；根据待测语音板的状态，通过对所述待测语音板播放所述音频测试文件，执行测试任务；生成用于指示所述待测语音板性能的测试结果的方式，通过根据待测语音板的状态进行自动语音性能测试、统计，达到了语音性能测试自动化、无人化操作的目的，从而实现了减少了测试人员、加快了测试速度的技术效果，进而解决了由于现有的语音测试方法需要多个测试人员进行长时间的念指令造成的测试耗时较长的技术问题。
附图说明
此处所说明的附图用来提供对本申请的进一步理解，构成本申请的一部分，本申请的示意性实施例及其说明用于解释本申请，并不构成对本申请的不当限定。在附图中：
图1是根据本申请实施例的一种可选的语音测试方法的流程示意图；
图2是根据本申请实施例的另一种可选的语音测试方法的流程示意图；
图3是根据本申请实施例的一种可选的软件界面示意图；
图4是根据本申请实施例的又一种可选的语音测试方法的流程示意图；
图5是根据本申请实施例的又一种可选的语音测试方法的流程示意图；
图6是根据本申请实施例的又一种可选的语音测试方法的流程示意图；
图7是根据本申请实施例的一种可选的语音测试装置的结构示意图。
具体实施方式
为了使本技术领域的人员更好地理解本申请方案，下面将结合本申请实施例中的附图，对本申请实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本申请一部分的实施例，而不是全部的实施例。基于本申请中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都应当属于 本申请保护的范围。
需要说明的是，本申请的说明书和权利要求书及上述附图中的术语“第一”、“第二”等是用于区别类似的对象，而不必用于描述特定的顺序或先后次序。应该理解这样使用的数据在适当情况下可以互换，以便这里描述的本申请的实施例能够以除了在这里图示或描述的那些以外的顺序实施。此外，术语“包括”和“具有”以及他们的任何变形，意图在于覆盖不排他的包含，例如，包含了一系列步骤或单元的过程、方法、系统、产品或设备不必限于清楚地列出的那些步骤或单元，而是可包括没有清楚地列出的或对于这些过程、方法、产品或设备固有的其它步骤或单元。
实施例1
根据本申请实施例，提供了一种语音测试方法的方法实施例，需要说明的是，在附图的流程图示出的步骤可以在诸如一组计算机可执行指令的计算机系统中执行，并且，虽然在流程图中示出了逻辑顺序，但是在某些情况下，可以以不同于此处的顺序执行所示出或描述的步骤。
图1是根据本申请实施例的语音测试方法，如图1所示，该方法包括如下步骤：
步骤S102，获取音频测试文件。
本实施例的语音测试方法，其测试需要依赖于音频测试文件、音频测试文件所对应的标注文件，音频测试文件包含有强制唤醒音频文件、强制停止播放音频文件、强制关闭语音音频文件和测试识别与唤醒所需要的样本音频文件；标注文件则是要测试的语音识别项目的音频文件所对应的信息(含测试音频文件名、测试音频文件的播放内容、测试音频文件所对应的格式信息)，这样就可以使计算机获取语音板通过串口所返回的信息与标注文件所含有的信息进行对比。
步骤S104，根据待测语音板的状态，通过对待测语音板播放音频测试文件，执行测试任务。
作为一种可选的实现方式，音频测试文件包括用于唤醒测试的第一音频文件(即唤醒词音频文件)，其中，根据待测语音板的状态，通过对待测语音板播放音频测试文件，执行测试任务包括：在待测语音板的状态为待唤醒状态的情况下，通过播放第一音频文件对待测语音板进行唤醒测试，得到待测语音板的唤醒率。
可选地，通过播放第一音频文件对待测语音板进行唤醒测试之前，方法还包括：获取待测语音板的状态；检测待测语音板的状态；若待测语音板的状态为待唤醒状态，触发执行过播放第一音频文件对待测语音板进行唤醒测试；若待测语音板的状态为唤 醒状态，通过对待测语音板播放第三音频文件，控制待测语音板退出唤醒状态；若待测语音板的状态为信源播放状态，通过对待测语音板播放第四音频文件，控制待测语音板停止播放信源，并在待测语音板停止播放信源之后，通过对待测语音板播放第三音频文件，控制待测语音板退出唤醒状态。
本实施例的语音测试方法，在测试一条语音指令之前，语音板必须处于待唤醒状态，方才符合实际用户使用的原则。自动化测试软件，首先需检测语音板是否处于待唤醒状态，若语音板的状态是唤醒状态，则让计算机通过高保真音箱对语音板播放强制关闭语音音频文件(即上述的第三音频文件)，以让其退出唤醒状态；若语音板处于信源播放状态，则让计算机通过高保真音箱对语音板播放强制停止播放音频文件(即上述的第四音频文件)，待获取到语音板停止播放信源的状态之后，再播放强制关闭语音音频文件(即上述的第三音频文件)，以让其退出唤醒状态。
可选地，通过播放第一音频文件对待测语音板进行唤醒测试，得到待测语音板的唤醒率包括：通过高保真音箱对待测语音板播放第一音频文件中的唤醒词；获取待测语音板返回的第一文字；判断第一文字是否为唤醒词；若第一文字为唤醒词，确定唤醒成功；根据唤醒成功的次数，得到唤醒率。
可选地，本实施例的语音测试方法还包括：若第一文字不为唤醒词，再次播放唤醒词；获取待测语音板返回的第二文字；若第二文字为唤醒词，确定唤醒成功。
可选地，本实施例的语音测试方法还包括：若第二文字不为唤醒词，重复播放用于强制唤醒待测语音板的第五音频；若在预设次数内成功唤醒待测语音板，继续进行唤醒测试；若超过预设次数未唤醒待测语音板，则结束唤醒测试。
本实施例测试语音板的唤醒率，需在语音板处于待唤醒状态进行。确保语音板在待唤醒状态，此时让计算机通过高保真音箱对语音板播放不同的唤醒词音频文(即上述的第一音频文件)件，等语音板通过串口返回获取到的音频信号所对应的文字之后，判断返回的文字内容是否为唤醒词，若是，则唤醒成功次数加1；若不是唤醒词，或者唤醒词不对，则再次重复播放当前音频文件1次；若成功，则唤醒成功次数加1，若唤醒不成功，则唤醒失败次数加1；当两次播放都唤不醒语音板，则重复播放强制唤醒音频文件(即上述的第五音频文件)，重复次数10次，只要10次内能唤醒语音板，则继续任务；当超过10次强唤醒时，则任务结束，软件提示“当前测试环境有误，请检查环境”。
作为另一种可选的实现方式，音频测试文件包括用于识别测试的第二音频文件及第二音频文件对应的标注文件，其中，根据待测语音板的状态，通过对待测语音板播 放音频测试文件，执行测试任务包括：在待测语音板的状态为唤醒状态的情况下，根据第二音频文件标注文件对待测语音板进行识别测试，得到待测语音板的识别率。
可选地，根据第二音频文件标注文件对待测语音板进行识别测试，得到待测语音板的识别率包括：通过高保真音箱对待测语音板播放第二音频文件中的指令词；获取待测语音板返回的文字内容；判断文字内容与标注文件中指示的信息是否相同；若文字内容与标注文件中指示的信息相同，确定识别成功；根据识别成功的次数，得到识别率。
可选地，方法还包括：若文字内容与标注文件中指示的信息不相同，再次播放指令词；若待测语音板再次返回的文字内容与标注文件中指示的信息相同，确定识别成功；若待测语音板再次返回的文字内容与标注文件中指示的信息不相同，确定识别失败。
本实施例测试识别率，必先唤醒语音板，当语音板被唤醒之后，方可进行识别率测试。识别率也有两次机会，高保真音箱对语音板播放不同的测试词音频文件(即上述的第二音频文件)，通过串口获取语音板返回的文字内容与标注文件里所对应的音频文件的播放内容、JSON(JavaScript Object Notation，一种轻量级的数据交换格式)是否一致，一致记为成功，不一致则记为失败；当第一次识别成功，则识别成功次数加1，进行下一个循环，若第1次不成功，则重复播放1次，当两次均失败，识别失败次数加1，再进行下一个循环。
可选地，标注文件包括以下至少之一：音频文件名、语音内容、格式。
步骤S106，生成用于指示待测语音板性能的测试结果。
把所有的任务完成后，软件会将其测试的唤醒成功次数转化成唤醒率，识别成功次数转换为识别率，生成最终的测试结果，而后输出至Excel表格。
通过此方法，实现了唤醒率和识别率同步进行，要测试不同人的声音对语音识别性能的影响，只需将不同人的声音信号录制成音频文件同时在标注文件里加入该音频文件名、对应的语音内容和JSON等信息即可。
通过上述步骤，可以通过根据待测语音板的状态进行自动语音性能测试、统计，达到了语音性能测试自动化、无人化操作的目的，从而实现了减少了测试人员、加快了测试速度的技术效果，进而解决了由于现有的语音测试方法需要多个测试人员进行长时间的念指令造成的测试耗时较长的技术问题。实现语音性能测试自动化、无人化操作；实现标准化测试，语音性能测试准确度高，测试现象可复现；缩短语音性能测试时间；排除人为主观因素的干扰，测试全过程标准化。
下面，如图2所示，对本实施例的语音测试方法进行说明：
步骤A，开始。
本实施例的语音测试方法使用高保真音箱播放音频文件来模拟人声(更逼真与人的说话声音)，将高保真音箱放置与1.5米高度的支架上，以模拟人的嘴巴位置。在确定好高保真音箱的位置后，打开软件进行测试，软件界面如图3所示。
步骤B，建立任务。
本实施例的软件可以基于Python2.7环境下运行，其测试需要依赖于测试音频文件、测试音频所对应的标注文件，其测试音频文件包含有强制唤醒音频文件、强制停止播放音频文件、强制关闭语音音频文件和测试识别与唤醒所需要的样本音频文件；标注文件则是要测试的语音识别项目的音频文件所对应的信息(含测试音频文件名、测试音频文件的播放内容、测试音频文件所对应的JSON信息)，这样就可以使计算机获取语音板通过串口所返回的信息与标注文件所含有的信息进行对比。软件进行测试是按照标注文件的内容进行的，有多少行就要循环测试多少次，每次作为一个任务进行。
步骤C，保证处于待唤醒状态。
在测试一条语音指令之前，语音板必须处于待唤醒状态，方才符合实际用户使用的原则。如图4所示，保证处于待唤醒状态的步骤包括：
步骤c1，开始。
步骤c2，获取语音板状态。
步骤c3，是否处于待唤醒状态。
若否，执行步骤c4；
若是，执行步骤c5。
步骤c4，播放退出唤醒的音频。
步骤c5，结束。
自动化测试软件，首先需检测语音板是否处于待唤醒状态，若语音板的状态是唤醒状态，则让计算机通过高保真音箱对语音板播放强制关闭语音音频文件，以让其退出唤醒状态；若语音板处于信源播放状态，则让计算机通过高保真音箱对语音板播放强制停止播放音频文件，待获取到语音板停止播放信源的状态之后，再播放强制关闭 语音音频文件，以让其退出唤醒状态。
步骤D，唤醒测试。
测试语音板的唤醒率，需在语音板处于待唤醒状态进行。如图5所示，唤醒测试的步骤包括：
步骤d1，开始。
步骤d2，播放唤醒词音频。
步骤d3，获取语音板状态。
步骤d4，是否被唤醒。
若否，执行d5；
若是，执行步骤d11。
步骤d5，再次播放唤醒词音频。
步骤d6，获取语音板状态。
步骤d7，是否被唤醒。
若否，执行步骤d8；
若是，执行步骤d11。
步骤d8，唤醒失败次数+1。
步骤d9，播放强唤醒词音频。
步骤d10，是否被唤醒。
若否，执行步骤d9；
若是，执行步骤d12。
步骤d11，唤醒成功次数+1。
步骤d12，结束。
经过C步骤后，确保语音板在待唤醒状态，此时让计算机通过高保真音箱对语音板播放不同的唤醒词音频文件，等语音板通过串口返回获取到的音频信号所对应的文字之后，判断返回的文字内容是否为唤醒词，若是，则唤醒成功次数加1；若不是唤醒词，或者唤醒词不对，则再次重复播放当前音频文件1次；若成功，则唤醒成功次 数加1，若唤醒不成功，则唤醒失败次数加1；当两次播放都唤不醒语音板，则重复播放强制唤醒音频文件，重复次数10次，只要10次内能唤醒语音板，则继续任务；当超过10次强唤醒时，则任务结束，软件提示“当前测试环境有误，请检查环境”。
步骤E，识别测试。
测试识别率，必先唤醒语音板，当语音板被唤醒之后，方可进行识别率测试。如图6所示，识别测试的步骤包括：
步骤e1，开始。
步骤e2，播放指令词音频。
步骤e3，获取语音板识别结果。
步骤e4，是否识别成功。
若否，执行步骤e5；
若是，执行步骤e9。
步骤e6，再次播放指令词音频。
步骤e7，获取语音板识别结果。
步骤e8，是否识别成功。
若是，执行步骤e9；
若否，执行步骤e10。
步骤e9，识别成功次数+1。
步骤e10，识别失败次数+1。
步骤e11，结束。
识别率也有两次机会，高保真音箱对语音板播放不同的测试词音频文件，通过串口获取语音板返回的文字内容与标注文件里所对应的音频文件的播放内容、JSON是否一致，一致记为成功，不一致则记为失败；当第一次识别成功，则识别成功次数加1，进行下一个循环，若第1次不成功，则重复播放1次，当两次均失败，识别失败次数加1，再进行下一个循环。
步骤F，任务是否完成。
步骤G，生成excel报表。
最后把所有的任务完成，软件会将其测试的唤醒成功次数转化成唤醒率，识别成功次数转换为识别率，而后输出至Excel表格。通过此方法，实现了唤醒率和识别率同步进行，要测试不同人的声音对语音识别性能的影响，只需将不同人的声音信号录制成音频文件同时在标注文件里加入该音频文件名、对应的语音内容和JSON等信息即可。
步骤H，结束。
本实施例的语音测试方法，采用GUI(Graphical User Interface，图形用户界面)界面，操作简单，易上手；利用计算机播放测试音频，通过串口接收语音板返回的数据从而进行分析，与标注文件进行对比，计算出语音识别测试结果，并将结果输出至Excel表，整个过程实现自动化；进而实现了唤醒率和识别率并行测试，大大缩短测试时间，提高测试效率，相对于传统语音性能测试，人员需求数量减少90％，测试时间减少50％。
实施例2
根据本申请实施例，提供了一种语音测试装置，如图7所示，该语音测试装置包括：
获取单元702，用于获取音频测试文件；测试单元704，用于根据待测语音板的状态，通过对所述待测语音板播放所述音频测试文件，执行测试任务；生成单元706，用于生成用于指示所述待测语音板性能的测试结果。
可选地，所述音频测试文件包括用于唤醒测试的第一音频文件，其中，测试单元704用于执行以下步骤根据待测语音板的状态，通过对所述待测语音板播放所述音频测试文件，执行测试任务：在所述待测语音板的状态为待唤醒状态的情况下，通过播放所述第一音频文件对所述待测语音板进行唤醒测试，得到所述待测语音板的唤醒率。
可选地，测试单元704，还用于获取所述待测语音板的状态；检测所述待测语音板的状态；若所述待测语音板的状态为所述待唤醒状态，触发执行所述过播放所述第一音频文件对所述待测语音板进行唤醒测试；若所述待测语音板的状态为唤醒状态，通过对所述待测语音板播放第三音频文件，控制所述待测语音板退出所述唤醒状态；若所述待测语音板的状态为信源播放状态，通过对所述待测语音板播放第四音频文件，控制所述待测语音板停止播放信源，并在所述待测语音板停止播放信源之后，通过对 所述待测语音板播放所述第三音频文件，控制所述待测语音板退出所述唤醒状态。
可选地，测试单元704用于执行以下步骤通过播放所述第一音频文件对所述待测语音板进行唤醒测试，得到所述待测语音板的唤醒率：通过高保真音箱对所述待测语音板播放所述第一音频文件中的唤醒词；获取所述待测语音板返回的第一文字；判断所述第一文字是否为所述唤醒词；若所述第一文字为所述唤醒词，确定唤醒成功；根据唤醒成功的次数，得到所述唤醒率。
可选地，测试单元704，还用于若所述第一文字不为所述唤醒词，再次播放所述唤醒词；获取所述待测语音板返回的第二文字；若所述第二文字为所述唤醒词，确定唤醒成功。
可选地，测试单元704，还用于若所述第二文字不为所述唤醒词，重复播放用于强制唤醒所述待测语音板的第五音频；若在预设次数内成功唤醒所述待测语音板，继续进行所述唤醒测试；若超过所述预设次数未唤醒所述待测语音板，则结束所述唤醒测试。
可选地，所述音频测试文件包括用于识别测试的第二音频文件及所述第二音频文件对应的标注文件，其中，测试单元704用于执行以下步骤根据待测语音板的状态，通过对所述待测语音板播放所述音频测试文件，执行测试任务：在所述待测语音板的状态为唤醒状态的情况下，根据所述第二音频文件所述标注文件对所述待测语音板进行识别测试，得到所述待测语音板的识别率。
可选地，测试单元704用于执行以下步骤根据所述第二音频文件所述标注文件对所述待测语音板进行识别测试，得到所述待测语音板的识别率：通过高保真音箱对所述待测语音板播放所述第二音频文件中的指令词；获取所述待测语音板返回的文字内容；判断所述文字内容与所述标注文件中指示的信息是否相同；若所述文字内容与所述标注文件中指示的信息相同，确定识别成功；根据识别成功的次数，得到所述识别率。
可选地，测试单元704，还用于若所述文字内容与所述标注文件中指示的信息不相同，再次播放所述指令词；若所述待测语音板再次返回的文字内容与所述标注文件中指示的信息相同，确定识别成功；若所述待测语音板再次返回的文字内容与所述标注文件中指示的信息不相同，确定识别失败。
可选地，所述标注文件包括以下至少之一：音频文件名、语音内容、格式。
根据本申请实施例，还提供了一种存储介质，所述存储介质包括存储的程序，其中，所述程序执行上述的语音测试方法。
根据本申请实施例，还提供了一种计算机终端，包括：存储器，用于存储程序；处理器，用于运行所述程序，其中，所述程序运行时执行上述的语音测试方法。
上述本申请实施例序号仅仅为了描述，不代表实施例的优劣。
在本申请的上述实施例中，对各个实施例的描述都各有侧重，某个实施例中没有详述的部分，可以参见其他实施例的相关描述。
在本申请所提供的几个实施例中，应该理解到，所揭露的技术内容，可通过其它的方式实现。其中，以上所描述的装置实施例仅仅是示意性的，例如所述单元的划分，可以为一种逻辑功能划分，实际实现时可以有另外的划分方式，例如多个单元或组件可以结合或者可以集成到另一个系统，或一些特征可以忽略，或不执行。另一点，所显示或讨论的相互之间的耦合或直接耦合或通信连接可以是通过一些接口，单元或模块的间接耦合或通信连接，可以是电性或其它的形式。
所述作为分离部件说明的单元可以是或者也可以不是物理上分开的，作为单元显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到多个单元上。可以根据实际的需要选择其中的部分或者全部单元来实现本实施例方案的目的。
另外，在本申请各个实施例中的各功能单元可以集成在一个处理单元中，也可以是各个单元单独物理存在，也可以两个或两个以上单元集成在一个单元中。上述集成的单元既可以采用硬件的形式实现，也可以采用软件功能单元的形式实现。
所述集成的单元如果以软件功能单元的形式实现并作为独立的产品销售或使用时，可以存储在一个计算机可读取存储介质中。基于这样的理解，本申请的技术方案本质上或者说对现有技术做出贡献的部分或者该技术方案的全部或部分可以以软件产品的形式体现出来，该计算机软件产品存储在一个存储介质中，包括若干指令用以使得一台计算机设备(可为个人计算机、服务器或者网络设备等)执行本申请各个实施例所述方法的全部或部分步骤。而前述的存储介质包括：U盘、只读存储器(ROM，Read-Only Memory)、随机存取存储器(RAM，Random Access Memory)、移动硬盘、磁碟或者光盘等各种可以存储程序代码的介质。
以上所述仅是本申请的优选实施方式，应当指出，对于本技术领域的普通技术人员来说，在不脱离本申请原理的前提下，还可以做出若干改进和润饰，这些改进和润饰也应视为本申请的保护范围。
工业实用性
本申请实施例提供的方案可应用于语音测试领域，在本申请实施例中，采用获取音频测试文件；根据待测语音板的状态，通过对所述待测语音板播放所述音频测试文件，执行测试任务；生成用于指示所述待测语音板性能的测试结果的方式，通过根据待测语音板的状态进行自动语音性能测试、统计，达到了语音性能测试自动化、无人化操作的目的，从而实现了减少了测试人员、加快了测试速度的技术效果。

Claims (13)

  1. 一种语音测试方法，包括：
    获取音频测试文件；
    根据待测语音板的状态，通过对所述待测语音板播放所述音频测试文件，执行测试任务；
    生成用于指示所述待测语音板性能的测试结果。
  2. 根据权利要求1所述的方法，其中，所述音频测试文件包括用于唤醒测试的第一音频文件，其中，根据待测语音板的状态，通过对所述待测语音板播放所述音频测试文件，执行测试任务包括：
    在所述待测语音板的状态为待唤醒状态的情况下，通过播放所述第一音频文件对所述待测语音板进行唤醒测试，得到所述待测语音板的唤醒率。
  3. 根据权利要求2所述的方法，其中，通过播放所述第一音频文件对所述待测语音板进行唤醒测试之前，所述方法还包括：
    获取所述待测语音板的状态；
    检测所述待测语音板的状态；
    若所述待测语音板的状态为所述待唤醒状态，触发执行所述过播放所述第一音频文件对所述待测语音板进行唤醒测试；
    若所述待测语音板的状态为唤醒状态，通过对所述待测语音板播放第三音频文件，控制所述待测语音板退出所述唤醒状态；
    若所述待测语音板的状态为信源播放状态，通过对所述待测语音板播放第四音频文件，控制所述待测语音板停止播放信源，并在所述待测语音板停止播放信源之后，通过对所述待测语音板播放所述第三音频文件，控制所述待测语音板退出所述唤醒状态。
  4. 根据权利要求2所述的方法，其中，通过播放所述第一音频文件对所述待测语音板进行唤醒测试，得到所述待测语音板的唤醒率包括：
    通过高保真音箱对所述待测语音板播放所述第一音频文件中的唤醒词；
    获取所述待测语音板返回的第一文字；
    判断所述第一文字是否为所述唤醒词；
    若所述第一文字为所述唤醒词，确定唤醒成功；
    根据唤醒成功的次数，得到所述唤醒率。
  5. 根据权利要求4所述的方法，其中，所述方法还包括：
    若所述第一文字不为所述唤醒词，再次播放所述唤醒词；
    获取所述待测语音板返回的第二文字；
    若所述第二文字为所述唤醒词，确定唤醒成功。
  6. 根据权利要求5所述的方法，其中，所述方法还包括：
    若所述第二文字不为所述唤醒词，重复播放用于强制唤醒所述待测语音板的第五音频；
    若在预设次数内成功唤醒所述待测语音板，继续进行所述唤醒测试；
    若超过所述预设次数未唤醒所述待测语音板，则结束所述唤醒测试。
  7. 根据权利要求1所述的方法，其中，所述音频测试文件包括用于识别测试的第二音频文件及所述第二音频文件对应的标注文件，其中，根据待测语音板的状态，通过对所述待测语音板播放所述音频测试文件，执行测试任务包括：
    在所述待测语音板的状态为唤醒状态的情况下，根据所述第二音频文件所述标注文件对所述待测语音板进行识别测试，得到所述待测语音板的识别率。
  8. 根据权利要求7所述的方法，其中，根据所述第二音频文件所述标注文件对所述待测语音板进行识别测试，得到所述待测语音板的识别率包括：
    通过高保真音箱对所述待测语音板播放所述第二音频文件中的指令词；
    获取所述待测语音板返回的文字内容；
    判断所述文字内容与所述标注文件中指示的信息是否相同；
    若所述文字内容与所述标注文件中指示的信息相同，确定识别成功；
    根据识别成功的次数，得到所述识别率。
  9. 根据权利要求8所述的方法，其中，所述方法还包括：
    若所述文字内容与所述标注文件中指示的信息不相同，再次播放所述指令词；
    若所述待测语音板再次返回的文字内容与所述标注文件中指示的信息相同，确定识别成功；
    若所述待测语音板再次返回的文字内容与所述标注文件中指示的信息不相同，确定识别失败。
  10. 根据权利要求7至9中任一项所述的方法，其中，所述标注文件包括以下至少之一：音频文件名、语音内容、格式。
  11. 一种语音测试装置，包括：
    获取单元，设置为获取音频测试文件；
    测试单元，设置为根据待测语音板的状态，通过对所述待测语音板播放所述音频测试文件，执行测试任务；
    生成单元，设置为生成指示所述待测语音板性能的测试结果。
  12. 一种存储介质，所述存储介质包括存储的程序，其中，所述程序执行权利要求1至10中任意一项所述的语音测试方法。
  13. 一种计算机终端，包括：
    存储器，设置为存储程序；
    处理器，设置为运行所述程序，其中，所述程序运行时执行权利要求1至10中任意一项所述的语音测试方法。
PCT/CN2018/118976 2018-03-07 2018-12-03 语音测试方法及装置 WO2019169914A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201810188414.X 2018-03-07
CN201810188414.XA CN108597494A (zh) 2018-03-07 2018-03-07 语音测试方法及装置

Publications (1)

Publication Number Publication Date
WO2019169914A1 true WO2019169914A1 (zh) 2019-09-12

Family

ID=63625785

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2018/118976 WO2019169914A1 (zh) 2018-03-07 2018-12-03 语音测试方法及装置

Country Status (2)

Country Link
CN (1) CN108597494A (zh)
WO (1) WO2019169914A1 (zh)

Families Citing this family (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108597494A (zh) * 2018-03-07 2018-09-28 珠海格力电器股份有限公司 语音测试方法及装置
CN111354335A (zh) * 2018-12-24 2020-06-30 深圳市优必选科技有限公司 一种语音识别测试方法、装置、存储介质及终端设备
CN113851109A (zh) * 2019-02-28 2021-12-28 百度在线网络技术（北京）有限公司 多音区唤醒测试方法、装置及存储介质
CN109817219A (zh) * 2019-03-19 2019-05-28 四川长虹电器股份有限公司 语音唤醒测试方法及系统
CN112309430A (zh) * 2019-07-31 2021-02-02 广东美的制冷设备有限公司 家电设备及其自检方法和装置
CN111179907A (zh) * 2019-12-31 2020-05-19 深圳Tcl新技术有限公司 语音识别测试方法、装置、设备及计算机可读存储介质
CN111341296B (zh) * 2020-02-17 2023-12-12 智达诚远科技有限公司 一种语音控制的响应测试方法、测试机和存储介质
CN113362806A (zh) * 2020-03-02 2021-09-07 北京奇虎科技有限公司 智能音响的评测方法、系统、存储介质及其计算机设备
CN113556661B (zh) * 2020-04-26 2023-04-07 阿里巴巴集团控股有限公司 电声检测与信息输出方法、系统、设备及存储介质
CN111611169A (zh) * 2020-05-22 2020-09-01 深圳市亿道数码技术有限公司 一种语音助手唤醒率自动化测试方法及测试工具
CN111739512A (zh) * 2020-06-18 2020-10-02 中汽院智能网联科技有限公司 一种基于实车的语音唤醒率测试方法、系统、设备及介质
CN111739513B (zh) * 2020-07-22 2020-12-11 江苏清微智能科技有限公司 自动化语音唤醒测试系统及其测试方法
CN111953764B (zh) * 2020-08-07 2023-04-07 杭州国芯科技股份有限公司 人工智能语音算法自动化测试方法
CN111933108B (zh) * 2020-09-25 2021-01-12 蘑菇车联信息科技有限公司 一种智能网联终端智能语音交互系统自动化测试方法

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102723080A (zh) * 2012-06-25 2012-10-10 惠州市德赛西威汽车电子有限公司 一种语音识别测试系统及方法
CN103578463A (zh) * 2012-07-27 2014-02-12 腾讯科技（深圳）有限公司 自动化测试方法及测试装置
CN104837010A (zh) * 2015-04-24 2015-08-12 青岛海信电器股份有限公司 语音遥控测试方法、装置及系统
CN106228986A (zh) * 2016-07-26 2016-12-14 北京奇虎科技有限公司 一种语音识别引擎的自动化测试方法、装置和系统
CN107516510A (zh) * 2017-07-05 2017-12-26 百度在线网络技术（北京）有限公司 一种智能设备自动化语音测试方法及装置
CN108597494A (zh) * 2018-03-07 2018-09-28 珠海格力电器股份有限公司 语音测试方法及装置

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104464755B (zh) * 2014-12-02 2018-01-16 科大讯飞股份有限公司 语音评测方法和装置
CN107221341A (zh) * 2017-06-06 2017-09-29 北京云知声信息技术有限公司 一种语音测试方法及装置

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102723080A (zh) * 2012-06-25 2012-10-10 惠州市德赛西威汽车电子有限公司 一种语音识别测试系统及方法
CN103578463A (zh) * 2012-07-27 2014-02-12 腾讯科技（深圳）有限公司 自动化测试方法及测试装置
CN104837010A (zh) * 2015-04-24 2015-08-12 青岛海信电器股份有限公司 语音遥控测试方法、装置及系统
CN106228986A (zh) * 2016-07-26 2016-12-14 北京奇虎科技有限公司 一种语音识别引擎的自动化测试方法、装置和系统
CN107516510A (zh) * 2017-07-05 2017-12-26 百度在线网络技术（北京）有限公司 一种智能设备自动化语音测试方法及装置
CN108597494A (zh) * 2018-03-07 2018-09-28 珠海格力电器股份有限公司 语音测试方法及装置

Also Published As

Publication number Publication date
CN108597494A (zh) 2018-09-28

Similar Documents

Publication Publication Date Title
WO2019169914A1 (zh) 语音测试方法及装置
CN107516510B (zh) 一种智能设备自动化语音测试方法及装置
US8990082B2 (en) Non-scorable response filters for speech scoring systems
CN107274906A (zh) 语音信息处理方法、装置、终端及存储介质
CN110457432A (zh) 面试评分方法、装置、设备及存储介质
CN110379410A (zh) 语音响应速度自动分析方法及系统
CN109065046A (zh) 语音唤醒的方法、装置、电子设备及计算机可读存储介质
US20150066504A1 (en) System and Method for Determining the Compliance of Agent Scripts
CN104464751A (zh) 发音韵律问题的检测方法及装置
CN110164474B (zh) 语音唤醒自动化测试方法及系统
CN104464757A (zh) 语音评测方法和语音评测装置
CN109215647A (zh) 语音唤醒方法、电子设备及非暂态计算机可读存储介质
CN110136748A (zh) 一种节奏识别校正方法、装置、设备及存储介质
CN109785683A (zh) 用于模拟口语考试现场的方法、装置、电子设备以及介质
US20090275005A1 (en) Methods, Systems, and Computer Program Products for Speech Assessment
CN111081260A (zh) 一种唤醒词声纹的识别方法及系统
JP7125042B2 (ja) 脳活動を利用した語学能力評価装置、及び語学能力評価システム
CN104299612A (zh) 模仿音相似度的检测方法和装置
KR102060229B1 (ko) 순차통역 자습 보조 방법 및 이를 수행하기 위한 기록매체
US20220215839A1 (en) Method for determining voice response speed, related device and computer program product
CN110503941B (zh) 语言能力评测方法、装置、系统、计算机设备及存储介质
CN110097874A (zh) 一种发音纠正方法、装置、设备以及存储介质
CN110085260A (zh) 一种单词音节重音识别校正方法、装置、设备以及介质
Dineley et al. Towards robust paralinguistic assessment for real-world mobile health (mHealth) monitoring: an initial study of reverberation effects on speech
Herms et al. CoLoSS: Cognitive load corpus with speech and performance data from a symbol-digit dual-task

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 18909241

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 18909241

Country of ref document: EP

Kind code of ref document: A1