A method, system, device and storage medium for voice recognition echo cancellation
Through software methods, identify and process audio data, convert voice into text and judge echo conditions, the echo cancellation problem of hardware circuit dependence is solved, flexible echo cancellation at any terminal, and reduce the cost of hardware circuit development.
Patent Information
- Application Number
- CN202211012307.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-08-23
- Publication Date
- 2024-12-10
- Estimated Expiration
- 2042-08-23
AI Technical Summary
The existing echo cancellation methods mainly rely on hardware circuits, and are complex in design and cannot be applied in existing microphone devices without echo cancellation functions, resulting in the inability to achieve echo cancellation of on-site radio equipment.
The audio data is identified and processed through software methods, convert the voice into text, and determine whether the echo conditions are met based on the text information and timestamp, and corresponding text replacement and deletion are performed to achieve echo cancellation.
It improves the flexibility of echo processing, can realize echo cancellation at any terminal, reduces the cost of hardware circuit development, and eliminates the need to modify existing equipment circuits.
Smart Images
Figure CN115376539B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of audio processing technology, and in particular to a method, system, device and storage medium for voice recognition echo cancellation. Background Art
[0002] When recording in an open indoor environment using a microphone or other speaker, sound will reflect within the confined space, causing an echo. This echo is then picked up and amplified by the microphone, making the person's spoken words inaudible. Existing echo cancellation is mostly implemented in hardware, using a specific echo cancellation circuit to eliminate the echo signal. However, this circuit design process is relatively complex. Furthermore, some echo cancellation circuits are already built into the microphone circuitry and cannot be used with existing microphones without echo cancellation capabilities, rendering existing live sound recording equipment unable to implement echo cancellation. Summary of the Invention
[0003] In order to overcome the deficiencies of the prior art, one of the objectives of the present invention is to provide a method for canceling echo in speech recognition, which eliminates echoes by software methods instead of hardware circuit development, thereby improving the flexibility of echo processing.
[0004] A second object of the present invention is to provide a speech recognition echo cancellation system.
[0005] A third object of the present invention is to provide an electronic device.
[0006] A fourth object of the present invention is to provide a computer-readable storage medium.
[0007] One of the purposes of the present invention is achieved by the following technical solution:
[0008] A speech recognition echo cancellation method, comprising:
[0009] Acquire audio data, identify and extract the voice information corresponding to the main voiceprint in the audio data, convert it into text, and mark each text with its corresponding reception timestamp;
[0010] It is determined whether the voice information meets the echo condition according to the text information corresponding to the main voiceprint and its timestamp. If the echo condition is met, the text information of the main voiceprint is extracted and the extracted text information is played.
[0011] Furthermore, when obtaining audio data, it also includes:
[0012] The microphone device that records the audio data is identified and marked as a target device, and the voiceprint information pre-bound to the target device is used as the main voiceprint.
[0013] Furthermore, after receiving the timestamp for each text tag, the following is also included:
[0014] Arranging each character in sequence according to the time sequence of the received timestamps and displaying them in each horizontally arranged main site on the real-time conversion interface; wherein each received timestamp corresponds to a unique main site;
[0015] When there are multiple characters with the same receiving timestamp, the characters with the same receiving timestamp are arranged vertically and displayed at the position corresponding to the receiving timestamp to form a secondary position until all voice information is converted and displayed.
[0016] Furthermore, before determining whether the echo condition is met:
[0017] The text at the primary site is subjected to phrase analysis. If the phrase analysis is abnormal, the text at the primary site and its corresponding secondary site are replaced until the text phrase analysis at the primary site is normal.
[0018] Furthermore, it also includes:
[0019] A modification instruction is received, and text on the primary site and its corresponding secondary site is replaced according to the modification instruction, or text on the primary site and / or the secondary site is replaced with specified text according to the modification instruction.
[0020] Furthermore, the method for determining whether the echo condition is met is:
[0021] Compare the text arranged at the primary site with the text arranged at the secondary site. If the similarity between the text arranged at the secondary site and the text arranged at the primary site reaches a preset value or higher, mark the text at the primary site that meets the comparison result as the primary target text, and mark the text at the secondary site that meets the comparison result as the secondary target text; and determine whether the reception timestamps of the primary target text and the secondary target text are within a preset time. If so, the echo condition is satisfied.
[0022] If the similarity between the text at the secondary site and the text at the primary site does not reach a preset value or the reception timestamps of the primary target text and the secondary target text exceed a preset time, the echo condition is not met.
[0023] Furthermore, if the echo condition is met, the method for extracting the text information of the main voiceprint is:
[0024] If the echo condition is met, the secondary target text and its corresponding voice arranged at the secondary site are deleted, and the text arranged at the main site is extracted and added to the playback file, and the text in the playback file is read aloud and played.
[0025] The second object of the present invention is achieved by adopting the following technical solution:
[0026] A speech recognition echo cancellation system, which executes the above-mentioned speech recognition echo cancellation method, comprises:
[0027] The audio conversion module is used to obtain audio data, identify and extract the voice information corresponding to the main voiceprint in the audio data, convert it into text, and mark each text with its corresponding reception timestamp;
[0028] The echo cancellation module is used to determine whether the voice information meets the echo condition based on the text information corresponding to the main voiceprint and its timestamp. If the echo condition is met, the text information of the main voiceprint is extracted and the extracted text information is played.
[0029] The third object of the present invention is achieved by adopting the following technical solution:
[0030] An electronic device comprises a processor, a memory and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned method for canceling echoes in speech recognition when executing the computer program.
[0031] The fourth object of the present invention is achieved by adopting the following technical solution:
[0032] A computer-readable storage medium stores a computer program, which implements the above-mentioned speech recognition echo cancellation method when executed.
[0033] Compared with the prior art, the present invention has the following beneficial effects:
[0034] After converting speech into text, the present invention determines whether there is an echo in the received audio data by identifying the text information and the reception time corresponding to the text. If repeated text appears within a preset time, it can be determined that the text is the text corresponding to the echo and can be eliminated. The present invention can achieve the purpose of echo cancellation at any terminal through software processing, replacing the traditional method of eliminating echo through hardware circuits, which can improve application flexibility, eliminate the need to modify the circuits of existing equipment, and reduce hardware circuit development costs. BRIEF DESCRIPTION OF THE DRAWINGS
[0035] Figure 1 Schematic diagram of the flow of the speech recognition echo cancellation method of the present invention;
[0036] Figure 2 Schematic diagram of site distribution in the visualization interface of the present invention;
[0037] Figure 3Schematic diagram of the modules of the speech recognition echo cancellation system of the present invention. DETAILED DESCRIPTION
[0038] The present invention will be further described below in conjunction with the accompanying drawings and specific implementation methods. It should be noted that, under the premise of no conflict, the various embodiments or technical features described below can be arbitrarily combined to form new embodiments.
[0039] Example 1
[0040] This embodiment provides a method for voice recognition echo cancellation. Before using this method, multiple microphone devices and at least one terminal device must be configured. The multiple microphone devices are used to record audio data from people speaking. Each person recording can pre-enter their voiceprint information. Before a meeting or speech, the administrator binds the recording person's voiceprint information to the microphone device at their location and marks the recording person's voiceprint information as the primary voiceprint. The microphone devices corresponding to the remaining silent persons remain muted or turned off to reduce noise. The terminal device is equipped with a visual interface, allowing administrators to control the on / off status of each microphone device. Administrators can also view the text corresponding to the recording person's speech content in real time in the visual interface. Administrators can also edit and play the text information, leveraging the data processing capabilities of the terminal device to achieve echo cancellation.
[0041] refer to Figure 1 As shown, the method specifically includes the following steps:
[0042] Step S1: Acquire audio data, identify and extract the voice information corresponding to the main voiceprint in the audio data and convert it into text, and mark each text with its corresponding reception timestamp;
[0043] When any recording person speaks, the recording person uses the pre-bound microphone device to record the audio data of the voice; when audio data is detected, the microphone device that records the audio data is identified and marked as the target device, and the voiceprint information pre-bound to the target device is used as the main voiceprint.
[0044] Then, the voiceprint features of each sound source in the audio data are identified, the voice information corresponding to the main voiceprint in the audio data is extracted and converted into text, and the corresponding reception timestamp is marked for each text according to the audio reception time corresponding to each text, and the reception timestamp is used as one of the bases for echo judgment.
[0045] In the visualization interface of this embodiment, there is a real-time conversion interface, in which there are multiple sites arranged in a matrix, that is, sites arranged in multiple rows horizontally and sites arranged in multiple columns vertically; and each site is numbered in order from top to bottom and from left to right.
[0046] When converting voice information into text, the text is arranged in sequence according to the time sequence of the reception timestamp of each text and displayed at each location, and each location where the text is written is marked as a primary location; wherein each reception timestamp corresponds to a unique primary location. Figure 2 As shown, Figure 2 The site in the row where A is located is the main site.
[0047] When there are multiple characters in the same receiving timestamp during the conversion process, one of the characters with the same receiving timestamp will be written to the primary site corresponding to the receiving timestamp, and the other characters with the same receiving timestamp will be written to the secondary sites corresponding to the primary site. Each primary site may have its corresponding secondary site, and the secondary sites are vertically distributed on one side of the corresponding primary site. Figure 2 As shown, Figure 2 In the middle, B is the secondary site of the main site A1.
[0048] Each recorded audio sentence is converted into text and the location is written as described above until all the voice information corresponding to the main voiceprint is converted and displayed.
[0049] Step S2: judging whether the voice information meets the echo condition based on the text information corresponding to the main voiceprint and its timestamp; if the echo condition is met, extracting the text information of the main voiceprint and playing the extracted text information.
[0050] Before the method to determine whether the echo condition is met:
[0051] A phrase analysis is performed on all the characters arranged at the main site, wherein the phrase analysis can be performed by dividing the characters into words and then comparing the divided words with existing phrases in a phrase library. If the same phrase exists in the phrase library, the divided words can be judged to be normal. If the same phrase is not found in the phrase library, the divided words can be judged to be abnormal.
[0052] If a phrase analysis anomaly occurs, the primary site with the anomaly is marked red. The text at the primary site marked red is replaced with the text at the corresponding secondary site. The replaced text is then re-segmented and phrase analysis is performed until the phrase analysis at the primary site is normal. When the phrase analysis is normal, the primary site's color switches back to green.
[0053] At the same time, the administrator can also manually modify the text content of each main site, that is, the administrator generates a modification instruction through a visual interface, and the system can replace the text of the main site and its corresponding secondary site according to the modification instruction, or replace the text on the main site and / or the secondary site with specified text according to the modification instruction.
[0054] After the above analysis and judgment, it is necessary to determine whether the echo condition is met based on the text on the main site. The method is as follows:
[0055] The text arranged at the main site and the text arranged at the secondary site are compared. If the similarity between the text arranged at the secondary site and the text arranged at the main site reaches a preset value or above, the text at the main site whose comparison result meets the standard is marked as the main target text, and the text at the secondary site whose comparison result meets the standard is marked as the secondary target text; and it is determined whether the receiving timestamps of the main target text and the secondary target text are within a preset time. If so, the echo condition is met.
[0056] If the similarity between the text at the secondary site and the text at the primary site does not reach a preset value, it means that the received repeated sound is not an echo; or if the receiving timestamps of the primary target text and the secondary target text exceed the preset time, it means that the received repeated sound does not conform to the echo time law. If any of the above situations exists, the echo condition is not met.
[0057] The preset time can be set in combination with the size of the indoor environment and the speed of sound return, and its specific value is not limited here.
[0058] If the echo condition is met, the method for extracting the text information of the main voiceprint is:
[0059] If the echo condition is met, the secondary target text and its corresponding voice at the secondary position are deleted, and the text at the primary position is extracted in sequence according to the primary position and added to the playback file, and the text in the playback file is read and played. If the echo condition is not met, it means that the current audio does not have an echo, and the audio data can be played directly.
[0060] Example 2
[0061] This embodiment provides a speech recognition echo cancellation system, which executes the speech recognition echo cancellation method as described in the first embodiment. Figure 3 As shown, the system includes:
[0062] The audio conversion module is used to obtain audio data, identify and extract the voice information corresponding to the main voiceprint in the audio data, convert it into text, and mark each text with its corresponding reception timestamp;
[0063] The echo cancellation module is used to determine whether the voice information meets the echo condition based on the text information corresponding to the main voiceprint and its timestamp. If the echo condition is met, the text information of the main voiceprint is extracted and the extracted text information is played.
[0064] After converting speech into text, the present invention determines whether there is an echo in the received audio data by identifying the text information and the reception time corresponding to the text. If repeated text appears within a preset time, it can be determined that the text is the text corresponding to the echo and can be eliminated. The present invention can achieve the purpose of echo cancellation at any terminal through software processing, replacing the traditional method of eliminating echo through hardware circuits, which can improve application flexibility, eliminate the need to modify the circuits of existing equipment, and reduce hardware circuit development costs.
[0065] In some embodiments, an electronic device is also provided, which includes a processor, a memory, and a computer program stored in the memory and runnable on the processor. When the processor executes the computer program, the speech recognition echo cancellation method in Example 1 is implemented.
[0066] In some embodiments, a computer-readable storage medium is further provided, on which a computer program is stored. When the computer program is executed, the above-mentioned speech recognition echo cancellation method is implemented.
[0067] The above-mentioned systems, devices and storage media and the methods in the aforementioned embodiments are based on multiple aspects of the same inventive concept. The implementation process of the method has been described in detail above, so those skilled in the art can clearly understand the structure and implementation process of the system, device and storage medium in this embodiment based on the above description. For the sake of brevity of the specification, they will not be repeated here.
[0068] The above embodiments are only preferred embodiments of the present invention and cannot be used to limit the scope of protection of the present invention. Any non-substantial changes and replacements made by technicians in this field on the basis of the present invention fall within the scope of protection required by the present invention.
Claims
1. A speech recognition echo cancellation method, It is characterized in that include: Acquire audio data, identify and extract the voice information corresponding to the main voiceprint in the audio data and convert it into text, and mark each text with its corresponding receiving timestamp; Determining whether the voice information meets the echo condition according to the text information corresponding to the main voiceprint and its timestamp, and if the echo condition is met, extracting the text information of the main voiceprint and playing the extracted text information; After receiving the timestamp for each text tag, the method further includes: Arrange each character in sequence according to the time sequence of the received timestamps and display them in each horizontally arranged main site of the real-time conversion interface; wherein each received timestamp corresponds to a unique main site; When there are multiple characters with the same receiving timestamp, the characters with the same receiving timestamp are arranged vertically and displayed at the position corresponding to the receiving timestamp to form a secondary position until all the voice information is converted and displayed; The method to determine whether the echo condition is met is: Compare the text arranged at the primary site and the text arranged at the secondary site. If the similarity between the text arranged at the secondary site and the text arranged at the primary site reaches a preset value or above, mark the text at the primary site that meets the comparison result as the primary target text, and mark the text at the secondary site that meets the comparison result as the secondary target text; and determine whether the reception timestamps of the primary target text and the secondary target text are within a preset time, and if so, the echo condition is satisfied; If the similarity between the text at the secondary site and the text at the primary site does not reach a preset value or the receiving timestamps of the primary target text and the secondary target text exceed a preset time, the echo condition is not met.
2. The method for canceling echo in speech recognition according to claim 1, It is characterized in that When getting audio data, it also includes: The microphone device that records the audio data is identified and marked as a target device, and the voiceprint information pre-bound to the target device is used as the main voiceprint.
3. The speech recognition echo cancellation method according to claim 1, It is characterized in that Before judging whether the echo condition is met: The words at the primary site are subjected to phrase analysis. If the phrase analysis is abnormal, the words at the primary site and the corresponding secondary site are replaced until the word phrase analysis at the primary site is normal.
4. The method for canceling echo in speech recognition according to claim 1, It is characterized in that Also includes: A modification instruction is received, and text is replaced between the primary site and the secondary site corresponding thereto according to the modification instruction, or text on the primary site and / or the secondary site is replaced with specified text according to the modification instruction.
5. The speech recognition echo cancellation method according to claim 1, It is characterized in that If the echo condition is met, the method for extracting the text information of the main voiceprint is: If the echo condition is met, the secondary target text and its corresponding voice arranged at the secondary site are deleted, and the text arranged at the main site is extracted and added to the playback file, and the text in the playback file is read aloud and played.
6. A speech recognition echo cancellation system, It is characterized in that The method for canceling the speech recognition echo according to any one of claims 1 to 5 is executed, wherein the system comprises: The audio conversion module is used to obtain audio data, identify and extract the voice information corresponding to the main voiceprint in the audio data and convert it into text, and mark each text with its corresponding receiving timestamp; The echo cancellation module is used to determine whether the voice information meets the echo condition according to the text information corresponding to the main voiceprint and its timestamp, and if the echo condition is met, extract the text information of the main voiceprint and play the extracted text information.
7. An electronic device, It is characterized in that It comprises a processor, a memory and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the speech recognition echo elimination method according to any one of claims 1 to 5 is implemented.
8. A computer-readable storage medium, It is characterized in that A computer program is stored thereon, and when the computer program is executed, the method for eliminating echo in speech recognition according to any one of claims 1 to 5 is implemented.