Voice processing method and device

Abstract

The present disclosure provides a speech processing method and device, the method: obtaining a syllable in the audio to be processed as the current speech segment, judging whether the current speech segment meets the audio segmentation condition and the interactive response condition, the audio segmentation condition is set based on the ASR recognition process, and the interactive response condition is set based on the NLP processing process and the matching degree of the NLP processing result and the feedback message. If the current speech segment does not meet the audio segmentation condition and the interactive response condition, the next syllable in the audio to be processed is merged into the current speech segment, and the step of judging whether the current speech segment meets the audio segmentation condition and the interactive response condition is performed until the target speech segment that meets the audio segmentation condition and the interactive response condition is determined. Output a feedback message that matches the target speech segment. The method greatly improves the audio segmentation efficiency, improves the reaction speed of robots or intelligent devices in human-computer communication, and improves the human-computer interaction experience.

Description

Technical Field

[0001] The present disclosure relates to the field of data processing technology, and in particular to a speech processing method and device. Background Art

[0002] With the development of artificial intelligence technology, human-computer communication is becoming more and more common in order to save human resources.

[0003] Taking the human-computer dialogue scenario as an example, after the user speaks, the robot or smart device will recognize the user's voice and respond based on the user's voice.

[0004] In the process of robot or smart device recognizing user voice, it is necessary to segment the user voice, that is, to determine whether the audio segment has ended. In the related art, it is usually necessary to wait for a preset time length (for example, 500 milliseconds) after recognizing a syllable. If no valid audio signal is detected within the preset time length, it can be determined that the syllable has ended, and the syllable after the end can be truncated into an audio segment. If a valid audio signal is detected within the preset time length, the syllable can be determined to be a continuous syllable, and processing continues until the end of the syllable is detected. Obviously, the current method of segmenting user voice requires waiting for at least a preset time length (for example, 500 milliseconds), which greatly reduces the efficiency of user voice processing and reduces the response speed of robots or smart devices in human-computer communication, resulting in a poor human-computer communication experience.

[0005] In summary, how to improve the response speed of robots or intelligent devices in human-computer communication has become a technical problem that needs to be solved urgently. Summary of the invention

[0006] The present disclosure provides a speech processing method and device for improving the response speed of a robot or intelligent device in human-machine communication.

[0007] According to a first aspect of an embodiment of the present disclosure, the present disclosure provides a speech processing method, including:

[0008] Get a syllable in the audio to be processed as the current voice segment;

[0009] Determine whether the current voice segment meets the audio segmentation conditions and interactive response conditions. The audio segmentation conditions are set based on the ASR recognition process, and the interactive response conditions are set based on the NLP processing process and the matching degree between the NLP processing result and the feedback message.

[0010] If the current voice segment does not meet the audio segmentation condition and the interactive response condition, then merge the next syllable in the audio to be processed into the current voice segment, and proceed to the step of determining whether the current voice segment meets the audio segmentation condition and the interactive response condition, until a target voice segment that meets the audio segmentation condition and the interactive response condition is determined;

[0011] Outputs a feedback message that matches the target speech segment.

[0012] According to a second aspect of an embodiment of the present disclosure, the present disclosure provides a speech processing device, including:

[0013] The processing module is configured to obtain a syllable in the audio to be processed as the current voice segment; determine whether the current voice segment meets the audio segmentation condition and the interactive response condition, the audio segmentation condition is set based on the ASR recognition process, and the interactive response condition is set based on the NLP processing process and the matching degree of the NLP processing result and the feedback message; if the current voice segment does not meet the audio segmentation condition and the interactive response condition, merge the next syllable in the audio to be processed into the current voice segment, and go to the step of determining whether the current voice segment meets the audio segmentation condition and the interactive response condition, until a target voice segment that meets the audio segmentation condition and the interactive response condition is determined;

[0014] The output module is configured to output a feedback message matching the target speech segment.

[0015] According to a third aspect of an embodiment of the present disclosure, an electronic device is provided, comprising a processor and a memory, wherein an executable code is stored on the memory, and when the executable code is executed by the processor, the processor can at least implement the speech processing method in the first aspect.

[0016] According to a fourth aspect of an embodiment of the present disclosure, a computer-readable storage medium is provided. When instructions in the computer-readable storage medium are executed by an electronic device, the electronic device is enabled to execute at least the speech processing method in the first aspect.

[0017] According to a fifth aspect of an embodiment of the present disclosure, there is provided a computer program product, comprising a computer program / instructions, which implement the speech processing method in the first aspect when executed by a processor.

[0018] The technical solution provided by the embodiments of the present disclosure brings at least the following beneficial effects:

[0019] In the present disclosure, for the audio to be processed, a syllable in the audio to be processed is obtained as the current voice segment, and it is determined whether the current voice segment meets the audio segmentation condition and the interactive response condition. The audio segmentation condition is set based on the ASR recognition process, and the interactive response condition is set based on the NLP processing process and the matching degree between the NLP processing result and the feedback message. Furthermore, if the current voice segment does not meet the audio segmentation condition and the interactive response condition, the next syllable in the audio to be processed is merged into the current voice segment, and the process proceeds to the step of determining whether the current voice segment meets the audio segmentation condition and the interactive response condition, until a target voice segment that meets the audio segmentation condition and the interactive response condition is determined. Finally, a feedback message matching the target voice segment is output. In the present disclosure, the audio segmentation conditions can be used to determine whether the current voice segment can successfully complete ASR recognition. For the voice segment that successfully completes ASR recognition, the interactive response conditions must be used to determine whether it contains complete interactive information sufficient to successfully implement NLP processing, thereby sending a matching feedback message for the voice segment that successfully implements NLP processing, avoiding the problem of reduced user voice segmentation efficiency and interaction efficiency due to waiting for at least a preset time in the related technology, greatly improving the audio segmentation efficiency, increasing the response speed of robots or smart devices in human-computer communication, and improving the human-computer interaction experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The drawings herein are incorporated into and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the description are used to explain the principles of the present disclosure, and do not constitute improper limitations on the present disclosure.

[0021] Figure 1 It is a schematic diagram showing a speech processing scenario according to an exemplary embodiment.

[0022] Figure 2 The figure is a flowchart of a speech processing method according to an exemplary embodiment.

[0023] Figure 3 The figure is a flowchart showing a judgment method according to an exemplary embodiment.

[0024] Figure 4 The figure is a schematic diagram showing the structure of a speech processing device according to an exemplary embodiment.

[0025] Figure 5 The diagram is a schematic structural diagram of an electronic device according to an exemplary embodiment. DETAILED DESCRIPTION

[0026] In order to enable ordinary persons in the art to better understand the technical solutions of the present disclosure, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the accompanying drawings.

[0027] It should be noted that the terms "first", "second", etc. in the specification and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of devices and methods consistent with some aspects of the present disclosure as detailed in the appended claims.

[0028] As mentioned above, in a human-computer dialogue scenario, after a user speaks, the robot or smart device will recognize the user's voice and respond based on the user's voice. In actual applications, the response to the user's voice can be in text form, voice form, or any other form.

[0029] In the process of robot or intelligent device recognizing user voice, it is necessary to segment the user voice, that is, to determine whether the audio segment has ended. In the related art, it is usually necessary to wait for a preset time length (for example, 500 milliseconds) after recognizing a syllable. If no valid audio signal is detected within the preset time length, it can be determined that the syllable has ended, and the syllable after the end can be truncated into an audio segment. If a valid audio signal is detected within the preset time length, it can be determined that the syllable is a continuous syllable, and the processing continues until the end of the syllable is detected. Obviously, the current method of segmenting user voice requires waiting for at least a preset time length (for example, 500 milliseconds), which greatly reduces the efficiency of user voice processing and reduces the reaction speed of robots or intelligent devices in human-computer communication, resulting in a poor human-computer communication experience. In summary, how to improve the reaction speed of robots or intelligent devices in human-computer communication has become a technical problem that needs to be solved urgently.

[0030] To solve at least one technical problem existing in the related art, the present disclosure provides a speech processing method and device.

[0031] The core idea of ​​the above technical solution is: to obtain voice segments from the audio to be processed, and to determine whether the currently obtained voice segment (i.e., syllable or syllable combination) can successfully complete automatic speech recognition (Automatic Speech Recognition, ASR) and implement NLP processing in combination with the audio segmentation conditions and interactive response conditions, so as to send a matching feedback message for the voice segment that already contains complete interactive information (such as questions raised by the user), thereby avoiding the problem of reduced user voice segmentation efficiency and interaction efficiency caused by waiting for at least a preset time in related technologies, greatly improving audio segmentation efficiency, improving the response speed of robots or smart devices in human-computer communication, and improving the human-computer interaction experience.

[0032] Taking voice information containing N syllables as an example, compared with the recognition method in the related art, this solution needs to call N system resources for judgment, which consumes N-1 (the number of syllables minus one) more system resources. However, since the process of obtaining the voice segment and the process of judging whether the voice segment meets the audio segmentation conditions can be carried out in parallel, the actual time consumed by this solution is the time consumed for obtaining N syllables and the time consumed for one judgment process. Obviously, it greatly improves the response speed of human-computer interaction and reduces the waiting time of human-computer interaction services.

[0033] In the present disclosure, the above scheme can be implemented by an electronic device, which can be a terminal device such as a robot, a mobile phone, a tablet computer, a wearable device (such as a smart bracelet, VR glasses, etc.), a PC, etc. Taking a robot as an example, a dedicated application installed in the robot can be called for implementation, other applications set in the robot can be called for implementation, and a cloud server can be called by the robot for implementation. Alternatively, the above scheme can also be implemented by a server. The above scheme can also be implemented by the cooperation of multiple electronic devices. For example, the server can send the execution result to the terminal device for the terminal device to display the execution result. The server can be a physical server containing an independent host, or it can also be a virtual server hosted by a host cluster, or it can also be a cloud server, which is not limited by the present disclosure.

[0034] by Figure 1 Taking the scenario shown as an example, the terminal device can transmit the recorded audio to be processed to the server side, and the server executes the above solution and returns the matching feedback message (i.e., the execution result) to the terminal device, so that the terminal device can display and play the corresponding feedback message. Figure 1 In the example, the terminal device can be one or more of a robot, a mobile phone, and a PC.

[0035] Based on the core idea introduced above, the present disclosure provides a speech processing method. Figure 2 FIG. 1 is a flow chart of a speech processing method provided by an exemplary embodiment of the present disclosure. Figure 2As shown, the method includes:

[0036] 201. Obtain a syllable in the audio to be processed as the current voice segment;

[0037] 202. Determine whether the current voice segment meets the audio segmentation condition and the interactive response condition;

[0038] 203. If the current voice segment does not meet the audio segmentation condition and the interactive response condition, merge the next syllable in the audio to be processed into the current voice segment, and proceed to the step of determining whether the current voice segment meets the audio segmentation condition and the interactive response condition, until a target voice segment that meets the audio segmentation condition and the interactive response condition is determined;

[0039] 204. Output a feedback message matching the target voice segment.

[0040] In the above method, the audio segmentation conditions and the interactive response conditions are used to determine whether the current voice segment (i.e., a syllable or a combination of syllables) extracted from the audio to be processed can successfully complete ASR recognition and implement NLP processing, thereby sending a matching feedback message for the voice segment that already contains complete interactive information (such as questions raised by the user), thereby avoiding the problem of reduced user voice segmentation efficiency and interaction efficiency caused by waiting for at least a preset time in the related technology, greatly improving the audio segmentation efficiency, increasing the response speed of robots or smart devices in human-computer communication, and improving the human-computer interaction experience.

[0041] The following describes various steps in the speech processing method in conjunction with specific embodiments.

[0042] In 201, a syllable in the audio to be processed is obtained as the current voice segment.

[0043] In the present disclosure, the audio to be processed includes but is not limited to the interactive voice of the user, such as questions, query requests, and records initiated by the user. Taking the robot as an example, the audio to be processed can be recorded by the sound receiving module (such as a microphone) carried by the robot. It is worth noting that in addition to audio, the voice clips in the present disclosure can also be obtained from video or other types of data.

[0044] Specifically, in 201, the starting position and the ending position of a syllable in the audio to be processed are determined according to the convergence trend of the vocal energy in the audio to be processed. Then, the corresponding audio segment is cut out from the audio to be processed as the current voice segment according to the starting position and the ending position. This step can be implemented by voice activity detection (VAD). In practical applications, voice activity detection is also called voice endpoint detection and voice boundary detection.

[0045] For example, after receiving the voice emitted by the user, the terminal device monitors the change trend of the vocal energy of the user's voice. Specifically, if it is monitored that the vocal energy in the user's voice shows a convergence trend, the starting position and the end position of a syllable can be determined according to the vocal energy value in the user's voice. For example, assuming that the vocal energy increases to a first preset energy value, the position corresponding to the first preset energy value can be set as the starting position of the syllable. Conversely, assuming that the vocal energy decreases to a second preset energy value, the position corresponding to the second preset energy value can be set as the end position of the syllable. In practical applications, the first preset energy value may be the same as or different from the second preset energy value, and the present disclosure does not limit it. Furthermore, according to the starting position and the end position, the corresponding audio segment is intercepted from the user's voice as the current voice segment. In fact, based on the starting position and the end position determined in the above manner, the interception range of the voice segment can be expanded, so that the current voice segment contains closed sounds, breath sounds, etc., to avoid omissions in the interception of the voice segment.

[0046] In 202, it is determined whether the current voice segment meets the audio segmentation condition and the interactive response condition. The audio segmentation condition is set based on the ASR recognition process, and the interactive response condition is set based on the NLP processing process and the matching degree between the NLP processing result and the feedback message.

[0047] It is understandable that the audio segmentation condition is used to determine whether the current voice segment can successfully complete ASR recognition, that is, whether the current voice segment is pronounced completely. For the current voice segment that can successfully complete ASR recognition, it is considered that the current voice segment meets the audio segmentation condition, that is, the current voice segment is pronounced completely. In practical applications, the audio segmentation condition can be set according to syllable features. For example, syllable features include but are not limited to vocal energy and audio pitch. In this case, it is also necessary to further determine whether the current voice segment contains complete interactive information sufficient to successfully implement NLP processing through interactive response conditions, and to determine whether the matching degree of the NLP processing result and the feedback message reaches a preset value. It is understandable that complete interactive information refers to semantic information that can express clear interactive intentions, such as semantic information that can express the intended query content, semantic information that can express the intended question content, etc. The recognizability of the current voice segment can be determined based on the NLP processing result, wherein the recognizability is, for example, whether the current voice segment can be recognized by natural language processing (Natural Language Processing, NLP). The matching degree between the NLP processing result and the feedback message further reflects whether there is a feedback message matching the current voice segment, that is, whether an interactive response with the current voice segment can be achieved. The specific implementation of the audio segmentation condition and the interactive response condition is shown in the following embodiment, which will not be described here.

[0048] In the present disclosure, if the current voice segment meets the audio segmentation conditions and the interactive response conditions, a feedback message matching the current voice segment is output. Optionally, the audio to be processed is a voice question issued by the user, and the feedback message matching the current voice segment is a reply message to the voice question. In an optional embodiment, the step of outputting the feedback message matching the current voice segment can be implemented using Text To Speech (TTS) technology, specifically, a text file is converted into a natural language voice stream in real time through a neural network, and the converted voice stream is played through a player.

[0049] In 203, if the current voice segment does not meet any of the audio segmentation conditions and the interactive response conditions, it means that the current voice segment is not pronounced completely or does not meet the conditions required to achieve interactive response (such as not containing complete interactive information or there is no feedback message matching it). In this case, the next syllable can be merged into the current voice segment, and the process proceeds to the step of determining whether the current voice segment meets the audio segmentation conditions and the interactive response conditions, until a target voice segment that meets the audio segmentation conditions and the interactive response conditions is determined.

[0050] In the present disclosure, optionally, if the current voice segment obtained after merging the next sequence syllables meets the audio segmentation condition, it means that the pronunciation of the current voice segment is complete and meets the conditions required to achieve interactive response. In this case, a feedback message matching the current voice segment can be directly output. The feedback message here is similar to the feedback message in the previous embodiment and will not be elaborated.

[0051] If the current voice segment does not meet the audio segmentation conditions, it means that the current voice segment is not pronounced completely or does not meet the conditions required to achieve interactive response. In this case, go to step 203 to merge the next syllable into the current voice segment, and determine whether the current voice segment meets the audio segmentation conditions and the interactive response conditions. The above steps are executed repeatedly until the target voice segment that meets the audio segmentation conditions and the interactive response conditions is determined.

[0052] It is worth noting that in actual applications, the acquisition process and judgment process of each syllable in the audio to be processed in step 203 can be executed synchronously, thereby avoiding the problem of audio segmentation taking too long due to waiting for a preset time after recording, reducing the time required for voice segmentation, and improving audio segmentation efficiency.

[0053] In 204, a feedback message matching the current voice segment that meets the audio segmentation condition is output. The feedback message here is similar to the feedback message in the previous embodiment, and will not be described in detail.

[0054] pass Figure 2In the voice processing method shown, the audio segmentation conditions and the interactive response conditions are used to determine whether the current voice segment (i.e., a syllable or a combination of syllables) extracted from the audio to be processed can successfully complete ASR recognition and implement NLP processing, thereby sending a matching feedback message for the voice segment that already contains complete interactive information (such as questions raised by the user), avoiding the problem in the related art of reduced user voice segmentation efficiency and interaction efficiency due to waiting for at least a preset time length, greatly improving the audio segmentation efficiency, increasing the response speed of robots or smart devices in human-computer communication, and improving the human-computer interaction experience.

[0055] In the above or following embodiments, the audio segmentation condition may include but is not limited to: the convergence trend of the vocalization energy of the current voice segment meets the set trend. The interactive response condition may include but is not limited to: the current voice segment contains complete interactive information, and the matching degree between the text data and the feedback message reaches a preset value.

[0056] Based on the above audio segmentation conditions, the present disclosure optionally further provides a judgment method for judging whether the current voice segment meets the audio segmentation conditions and the interactive response conditions. Figure 3 A flowchart of a determination method provided by an exemplary embodiment of the present disclosure is provided, and the method comprises the following steps:

[0057] In 301, it is determined whether the convergence trend of the utterance energy of the current speech segment meets the set trend.

[0058] Optionally, the set trend is to gradually decrease to a preset vocalization energy threshold. Specifically, determine whether the vocalization energy of the current voice segment shows a decreasing trend and decreases to a preset vocalization energy threshold. If the vocalization energy of the current voice segment shows a decreasing trend and decreases to the vocalization energy threshold, then it is determined that the vocalization energy convergence trend of the current voice segment meets the set trend. If the vocalization energy of the current voice segment does not show a decreasing trend, or does not decrease to the vocalization energy threshold, then it is determined that the vocalization energy convergence trend of the current voice segment does not meet the set trend. In actual applications, the decreasing trend can be an overall change trend within a preset time length, during which certain fluctuations are allowed, which are not limited here.

[0059] If the convergence trend of the vocalization energy of the current voice segment meets the set trend, in this case, it means that the pronunciation of the current voice segment is complete. On this basis, it can be further detected whether the current voice segment contains complete interactive information.

[0060] If the convergence trend of the vocalization energy of the current voice segment does not meet the set trend, it is determined that the current voice segment does not meet the audio segmentation condition. In this case, it means that the pronunciation of the current voice segment is incomplete, so it is necessary to continue recording subsequent voice segments. Optionally, if the convergence trend of the vocalization energy does not meet the set trend, the next syllable is merged into the current voice segment. Note that, unlike the current voice segment before synthesis, the newly synthesized current voice segment here contains the next syllable.

[0061] Therefore, the vocalization energy of the current voice segment can be used to quickly detect whether the current voice segment is pronounced completely. If the current voice segment is not pronounced completely, continue to record subsequent voice segments to improve the efficiency of audio segmentation.

[0062] In 302, if the convergence trend of the vocalization energy meets the set trend, the current voice segment is converted into text data by using audio recognition processing. Specifically, the current voice segment can be converted into text by using Automatic Speech Recognition (ASR) technology, which provides a basis for subsequent audio segmentation operations.

[0063] If the current voice segment conversion fails, it means that the current voice segment does not contain a complete pronunciation or there is an error in the recording. In this case, it is determined that the current voice segment does not meet the audio segmentation conditions and needs to continue recording subsequent segments. In practical applications, optionally, if the current voice segment conversion fails, it can be determined that the current voice segment does not meet the interactive response conditions and the step of merging the next syllable into the current voice segment is performed.

[0064] In 303, it is determined whether the current voice segment is successfully converted. Specifically, determining whether the current voice segment is successfully converted is actually determining whether the current voice segment is recognizable.

[0065] If the current voice segment is successfully converted, it means that the current voice segment already contains complete semantic information. In this case, in 304, a feedback message matching the text data is obtained through natural language processing (Natural Language Processing, NLP).

[0066] Then, in 305, it is determined whether the matching degree between the text data and the feedback message reaches a preset value. The matching degree here can also be referred to as confidence level or confidence value. Specifically, it is determined whether the text data can be matched to the corresponding feedback message, that is, whether the converted text data is interactive. In simple terms, it is determined whether there is a feedback message that matches the current NLP processing result.

[0067] If the matching degree between the text data and the feedback message does not reach the preset value, it is determined that the current voice segment does not meet the interactive response conditions. In this case, it means that the voice information contained in the current voice segment is not complete interactive information, so it is necessary to continue recording subsequent segments. For example, suppose the question asked by the user is "What is the weather in Beijing?", and the semantic information contained in the current voice segment is "Beijing". In this case, since the word "Beijing" does not contain interactive intent, it is difficult to match a logical feedback message for "Beijing", that is, the matching degree between the matched feedback message and "Beijing" cannot reach the preset value, so the current situation can be determined as a match failure. Therefore, if the matching degree between the text data and the feedback message does not reach the preset value, go to the step of merging the next syllable into the current voice segment.

[0068] Of course, in actual applications, in addition to the above reasons, the reason why the matching degree between text data and feedback message does not reach the preset value may also be that the database storing feedback message does not include relevant questions, feedback message search times out, etc.

[0069] In 306, if the matching degree between the text data and the feedback message reaches a preset value, it means that the current voice segment already contains complete interactive information and meets the conditions for implementing interactive response. In this case, it can be determined that the current voice segment meets the audio segmentation conditions and the interactive response conditions.

[0070] Figure 3 In the judgment method shown, through the combination of audio segmentation conditions and interactive response conditions, a multi-dimensional assessment is made as to whether the current voice segment is pronounced completely, contains complete interactive information, and meets the conditions for achieving interactive response, thereby providing a basis for the audio segmentation operation and the generation process of feedback messages, helping to further improve the audio segmentation efficiency, increase the reaction speed of robots or smart devices in human-computer communication, and improve the human-computer interaction experience.

[0071] Figure 4 A speech processing device is provided in an embodiment of the present disclosure. Figure 4 As shown, the speech processing device includes:

[0072] The processing module 401 is configured to obtain a syllable in the audio to be processed as the current voice segment; determine whether the current voice segment meets the audio segmentation condition and the interactive response condition, the audio segmentation condition is set based on the ASR recognition process, and the interactive response condition is set based on the NLP processing process and the matching degree of the NLP processing result and the feedback message; if the current voice segment does not meet the audio segmentation condition and the interactive response condition, merge the next syllable in the audio to be processed into the current voice segment, and go to the step of determining whether the current voice segment meets the audio segmentation condition and the interactive response condition, until the target voice segment that meets the audio segmentation condition and the interactive response condition is determined;

[0073] The output module 402 is configured to output a feedback message matching the target voice segment.

[0074] Optionally, the output module 402 is further configured to: if the current voice segment meets the audio segmentation condition and the interactive response condition, output a feedback message matching the current voice segment.

[0075] Optionally, in the process of obtaining a syllable in the audio to be processed as the current voice segment, the processing module 401 is configured to:

[0076] Determine the starting position and ending position of a syllable in the audio to be processed according to the convergence trend of the vocalization energy in the audio to be processed;

[0077] According to the starting position and the ending position, a corresponding audio segment is cut out from the audio to be processed as the current voice segment.

[0078] Optionally, in the process of determining whether the current voice segment meets the audio segmentation condition and the interactive response condition, the processing module 401 is configured to:

[0079] Determine whether the vocal energy convergence trend of the current speech segment meets the set trend;

[0080] If the convergence trend of the vocalization energy meets the set trend, the current voice segment is converted into text data using audio recognition processing;

[0081] If the current voice segment is converted successfully, a feedback message matching the text data is obtained through natural language processing (NLP);

[0082] Determine whether the matching degree between the text data and the feedback message reaches a preset value;

[0083] If the matching degree between the text data and the feedback message reaches a preset value, it is determined that the current voice segment meets the audio segmentation condition.

[0084] Optionally, the processing module 401 is further configured to:

[0085] If the convergence trend of the vocalization energy does not conform to the set trend, it is determined that the current voice segment does not conform to the audio segmentation condition; or

[0086] If the current voice segment conversion fails, it is determined that the current voice segment does not meet the interactive response condition; or

[0087] If the matching degree between the text data and the feedback message does not reach a preset value, it is determined that the current voice segment does not meet the interactive response condition.

[0088] Optionally, the audio to be processed is a voice question posed by a user, and the feedback message is a reply message to the voice question.

[0089] The above-mentioned speech processing device can execute the systems or methods provided in the aforementioned embodiments. For the parts not described in detail in this embodiment, reference can be made to the relevant descriptions of the aforementioned embodiments, and they will not be repeated here.

[0090] In a possible design, the structure of the above-mentioned speech processing device can be implemented as an electronic device. Figure 5 As shown, the electronic device may include: a processor 21 and a memory 22. The memory 22 stores executable codes, and when the executable codes are executed by the processor 21, the processor 21 can at least implement the speech processing method provided in the above embodiments.

[0091] The structure of the electronic device may further include a communication interface 23 for communicating with other devices or a communication network.

[0092] In addition, the present disclosure also provides a computer-readable storage medium including instructions, on which executable codes are stored, and when the executable codes are executed by the processor of the wireless router, the processor executes the feature data processing method based on the neural network provided in the aforementioned embodiments. Optionally, the computer-readable storage medium can be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, etc.

[0093] In an exemplary embodiment, a computer program product is also provided, including a computer program / instruction, which, when executed by a processor, implements the feature data processing method based on a neural network provided in the aforementioned embodiments. The computer program / instruction is implemented by a program running on a terminal or a server.

[0094] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the invention disclosed herein. The present disclosure is intended to cover any variations, uses or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary techniques in the art that are not disclosed in the present disclosure. The description and examples are to be considered exemplary only, and the true scope and spirit of the present disclosure are indicated by the following claims.

[0095] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A speech processing method, characterized in that: include: Get a syllable in the audio to be processed as the current voice segment; Determine whether the current voice segment meets the audio segmentation condition and the interactive response condition, wherein the audio segmentation condition is set based on the ASR recognition process, and the audio segmentation condition is used to determine whether the current voice segment can successfully complete the ASR recognition, and the interactive response condition is set based on the NLP processing process and the matching degree between the NLP processing result and the feedback message, and the interactive response condition is used to determine whether the current voice segment is sufficient to successfully implement the NLP processing, and determine whether the matching degree between the NLP processing result and the feedback message reaches a preset value; If the current voice segment does not meet the audio segmentation condition and the interactive response condition, then merge the next syllable in the to-be-processed audio into the current voice segment, and proceed to the step of determining whether the current voice segment meets the audio segmentation condition and the interactive response condition, until a target voice segment that meets the audio segmentation condition and the interactive response condition is determined; Outputting a feedback message matching the target speech segment.

2. The method according to claim 1, characterized in that Also includes: If the current voice segment meets the audio segmentation condition and the interactive response condition, a feedback message matching the current voice segment is output.

3. The method according to claim 1, characterized in that Get a syllable in the audio to be processed as the current voice segment, including: Determining a starting position and an ending position of a syllable in the audio to be processed according to a convergence trend of the vocalization energy in the audio to be processed; According to the starting position and the ending position, a corresponding audio segment is cut out from the audio to be processed as the current voice segment.

4. The method according to claim 1, characterized in that: The step of determining whether the current voice segment meets the audio segmentation condition and the interactive response condition comprises: Determining whether the utterance energy convergence trend of the current voice segment meets the set trend; If the convergence trend of the vocalization energy meets the set trend, converting the current voice segment into text data by using audio recognition processing; If the current voice segment is converted successfully, a feedback message matching the text data is obtained through natural language processing (NLP); Determining whether the matching degree between the text data and the feedback message reaches a preset value; If the matching degree reaches the preset value, it is determined that the current voice segment meets the audio segmentation condition and the interactive response condition.

5. The method according to claim 4, characterized in that Also includes: If the convergence trend of the vocalization energy does not conform to the set trend, determining that the current voice segment does not conform to the audio segmentation condition; or If the conversion of the current voice segment fails, determining that the current voice segment does not meet the interactive response condition; or If the matching degree does not reach the preset value, it is determined that the current voice segment does not meet the interactive response condition.

6. The method according to any one of claims 1 to 5, characterized in that: The audio to be processed is a voice question uttered by a user, and the feedback message is a reply message to the voice question.

7. A speech processing device, characterized in that: include: A processing module is configured to obtain a syllable in the audio to be processed as a current voice segment; Determine whether the current voice segment meets the audio segmentation condition and the interactive response condition, wherein the audio segmentation condition is set based on the ASR recognition process, and the audio segmentation condition is used to determine whether the current voice segment can successfully complete the ASR recognition, and the interactive response condition is set based on the NLP processing process and the matching degree between the NLP processing result and the feedback message, and the interactive response condition is used to determine whether the current voice segment is sufficient to successfully implement the NLP processing, and determine whether the matching degree between the NLP processing result and the feedback message reaches a preset value; If the current voice segment does not meet the audio segmentation condition and the interactive response condition, then merge the next syllable in the to-be-processed audio into the current voice segment, and proceed to the step of determining whether the current voice segment meets the audio segmentation condition and the interactive response condition, until a target voice segment that meets the audio segmentation condition and the interactive response condition is determined; The output module is configured to output a feedback message matching the target voice segment.

8. An electronic device, characterized in that: include: processor; a memory for storing instructions executable by the processor; The processor is configured to execute the instructions to implement the speech processing method according to any one of claims 1 to 6.

9. A computer-readable storage medium, when instructions in the computer-readable storage medium are executed by an electronic device, the electronic device is enabled to execute the speech processing method according to any one of claims 1 to 6.

10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the speech processing method according to any one of claims 1 to 6 is implemented.

Images