Communication device control method, communication device and storage medium

Abstract

The present application discloses a control method for a call device, a call device and a storage medium. The present application relates to the field of voice recognition technology. If the method detects the establishment of a voice conversation, the voice data corresponding to the voice conversation is collected through the voice recognition chip. If a preset keyword is recognized in the voice data, a recording file corresponding to the voice conversation is generated. The recording file is then uploaded to a cloud server through the voice gateway, and the recognition result fed back by the cloud server is received. Finally, the processing action corresponding to the recognition result is executed. The above method collects voice data through a voice recognition chip for preliminary processing and judgment, and recognizes the recording file of the voice conversation in combination with the cloud server, thereby improving the accuracy of abnormal call recognition and reducing the processing cost of intelligent voice.

Description

Technical Field

[0001] The present application relates to the field of speech recognition technology, and in particular to a control method of a communication device, a communication device and a storage medium. Background Art

[0002] At present, the traditional solution for identifying abnormal calls is to mark phone numbers that have been reported too many times as abnormal calls and save the marked phone numbers. Therefore, when a call is received from such a marked number, the user is reminded through the label that the call is an abnormal call. In this way, when the object of the abnormal call generates multiple different virtual numbers through software, it is impossible to identify whether the virtual number is an abnormal call, resulting in low accuracy in abnormal call identification.

[0003] With the development and application of AI technology, AI-based abnormal call recognition technology has also emerged. However, it requires real-time or recorded intelligent detection of every call, which consumes a lot of computing power and has high processing costs.

[0004] The above contents are only used to assist in understanding the technical solution of the present application and do not constitute an admission that the above contents are prior art. Summary of the invention

[0005] The present application provides a control method for a call device, a call device and a storage medium, aiming to solve the problems of low accuracy in abnormal call identification and high identification cost in related solutions.

[0006] To achieve the above-mentioned purpose, the present application provides a control method for a communication device, which is applied to the communication device. The communication device is provided with a voice recognition chip and is connected to a cloud server through a voice gateway. The control method for the communication device includes the following steps:

[0007] If it is detected that a voice conversation is established, voice data corresponding to the voice conversation is collected by the voice recognition chip;

[0008] If a preset keyword is recognized in the voice data, a recording file corresponding to the voice conversation is generated;

[0009] Uploading the recording file to the cloud server through the voice gateway, and receiving the recognition result fed back by the cloud server;

[0010] Execute the processing action corresponding to the recognition result.

[0011] In one embodiment, if a preset keyword is recognized in the voice data, the step of generating a recording file corresponding to the voice conversation includes:

[0012] The signal processing module performs noise reduction and enhancement processing on the collected voice data to obtain target voice data;

[0013] Detecting sentence endpoints in the target voice data based on voice activation detection technology and extracting corresponding voice features;

[0014] Perform forward reasoning based on the sentence endpoints and the speech features to obtain a posterior probability distribution, and determine keywords in the speech data based on the posterior probability distribution;

[0015] When the keyword matches the preset keyword, a recording file corresponding to the voice conversation is generated.

[0016] In one embodiment, when the keyword matches a preset keyword, the step of generating a recording file corresponding to the voice conversation includes:

[0017] Determining a recognition result in a decoding search space according to the keyword;

[0018] Determining the accuracy of the recognition result according to the keyword, and when the accuracy is higher than a preset threshold, determining that the keyword matches a preset keyword;

[0019] Start recording and record the voice conversation to generate the recording file.

[0020] In one embodiment, the step of executing the processing action corresponding to the recognition result includes:

[0021] Obtaining a semantic recognition result and an intention recognition result from the recognition result;

[0022] Determining whether the semantic recognition result and the intention recognition result match according to a preset abnormal phone feature library;

[0023] When the semantic recognition result and the intention recognition result match the features in the abnormal phone feature library, determining that the phone corresponding to the recognition result is an abnormal phone;

[0024] Execute the processing action corresponding to the abnormal call result.

[0025] In one embodiment, the step of executing the processing action corresponding to the recognition result that the call is abnormal includes:

[0026] Receiving a call termination instruction sent by the voice gateway to terminate the call process corresponding to the voice session;

[0027] Alternatively, receiving a prompt instruction sent by the voice gateway, and starting a voice prompt function according to the prompt instruction;

[0028] Alternatively, obtaining an associated target phone through the voice gateway, and sending the prompt voice to the target phone through a call;

[0029] The telephone number corresponding to the voice conversation is identified, and the voiceprint data in the recording file is determined, and the telephone number and the voiceprint data are associated and stored in a blacklist database.

[0030] In one embodiment, the method further includes:

[0031] Identifying the voiceprint of the recording file through the voice gateway, and comparing the voiceprint with a pre-stored voiceprint blacklist library;

[0032] When there is a target voiceprint matching the voiceprint in the voiceprint blacklist library, determining that the target phone corresponding to the voiceprint is an abnormal phone;

[0033] Execute the corresponding processing action for abnormal calls.

[0034] In one embodiment, after the step of generating a recording file corresponding to the voice conversation if a preset keyword is recognized in the voice data, the method further includes:

[0035] Inputting the recording file into a voice imitation recognition model, and determining whether the recording file is AI synthesized speech according to the voice imitation recognition model;

[0036] Extracting the timbre parameters of the recording file according to the sound imitation recognition model, and comparing the timbre parameters with pre-stored target timbre parameters;

[0037] When the timbre parameters match the target timbre parameters, the recording file is determined to be AI synthesized speech, and a corresponding voice prompt is output.

[0038] In one embodiment, after the step of executing the processing action corresponding to the recognition result by the voice gateway, the method further includes:

[0039] Upon receiving a misjudged instruction fed back by a user, determining a misjudged phone number corresponding to the misjudged instruction;

[0040] Searching the blacklist database for the target number that matches the misjudged phone number and the associated target voiceprint data;

[0041] The target number and the target voiceprint data are deleted from the blacklist database.

[0042] In addition, to achieve the above-mentioned purpose, the present application provides a communication device, which includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the control method of the communication device as described above.

[0043] In addition, to achieve the above-mentioned purpose, the present application provides a storage medium, which is a computer-readable storage medium, and a computer program is stored on the computer-readable storage medium. When the computer program is executed by a processor, the steps of the control method of the communication device as described above are implemented.

[0044] The present application provides a control method for a call device, a call device and a storage medium. If a voice conversation is detected, the voice data corresponding to the voice conversation is collected through the voice recognition chip. If a preset keyword is recognized in the voice data, a recording file corresponding to the voice conversation is generated. The recording file is then uploaded to a cloud server through the voice gateway, and the recognition result fed back by the cloud server is received. Finally, the processing action corresponding to the recognition result is executed. The present application collects voice data through a voice recognition chip for preliminary processing and judgment, and recognizes the recording file of the voice conversation in combination with the cloud server, thereby improving the accuracy of abnormal call recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0046] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0047] Figure 1 A flowchart of a first embodiment of a control method for a communication device of the present application;

[0048] Figure 2 A flowchart of a second embodiment of the control method of a communication device of the present application;

[0049] Figure 3 A flowchart of a third embodiment of a method for controlling a communication device of the present application;

[0050] Figure 4 A flowchart of a fourth embodiment of a control method for a communication device of the present application;

[0051] Figure 5 It is a schematic diagram of the architecture of the hardware operating environment of the communication device involved in the embodiment of the present application.

[0052] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0053] It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0054] In order to better understand the above technical solution, exemplary embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present application are shown in the accompanying drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present application and to fully convey the scope of the present application to those skilled in the art.

[0055] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0056] The main solution of the present application is: if it is detected that a voice session is established, voice data corresponding to the voice session is collected by the voice recognition chip;

[0057] If a preset keyword is recognized in the voice data, a recording file corresponding to the voice conversation is generated;

[0058] Uploading the recording file to the cloud server through the voice gateway, and receiving the recognition result fed back by the cloud server;

[0059] Execute the processing action corresponding to the recognition result.

[0060] At present, the traditional solution for identifying abnormal calls is to mark phone numbers that have been reported too many times as abnormal calls and save the marked phone numbers. Therefore, when a call is received from such a marked number, the user is reminded through the label that the call is an abnormal call. In this way, when the object of the abnormal call generates multiple different virtual numbers through software, it is impossible to identify whether the virtual number is an abnormal call, resulting in low accuracy in abnormal call identification.

[0061] This application collects voice data through a voice recognition chip for preliminary processing and judgment, and combines with a cloud server to identify the recorded files of voice conversations, thereby improving the accuracy of abnormal call identification.

[0062] Embodiment 1

[0063] Please refer to Figure 1 In a first embodiment, the control method of the communication device comprises the following steps:

[0064] Step S10: If it is detected that a voice conversation is established, voice data corresponding to the voice conversation is collected by the voice recognition chip.

[0065] In this embodiment, the processing action is performed by a call device, which is provided with a voice recognition chip and is connected to a cloud server through a voice gateway. When the call device detects that a voice conversation is established, that is, it detects that the call device has received a call request, and the call device can be confirmed to have received the call request by detecting a change in the state of the telephone line. It should be noted that the call device can be a landline, a mobile phone or a telephone watch, and any call device that can make voice calls, and there is no limitation here. The voice recognition chip can be set in the call device, or on a device connected to the call device, and can collect voice data of the call device when conducting a voice conversation in real time.

[0066] It should be noted that when the establishment of a voice session is detected, the corresponding acquisition prompt information is fed back to the call device, that is, the collection permission for collecting voice data is requested from the user. The user receives the acquisition prompt information based on the front end of the call device, and clicks to agree or refuse in the front end interface. When the user clicks the control of agreement, it represents the user to authorize the voice recognition chip, that is, the voice recognition has the collection permission. At this time, the voice recognition chip executes the collection of voice data corresponding to the voice session. Otherwise, the voice recognition chip cannot execute the above-mentioned collection of voice data corresponding to the voice session. Alternatively, in another optional application scenario, the guardian of the user corresponding to the call device performs the authorization action. After the call device is started, it requests permission from the terminal corresponding to the guardian, and the guardian authorizes the permission to the call device, so that the voice chip in the call device has the permission to collect voice data.

[0067] Before collecting the voice data of the communication device through the voice recognition chip, it is also necessary to ensure that the voice recognition chip built into the communication device functions normally and is correctly connected to the audio input terminal of the communication device. This can be done by checking whether the driver and algorithm library of the voice recognition chip have been correctly installed and configured. Start the software part of the voice recognition chip and ensure that the voice recognition chip is in standby mode. Then configure the parameters of the voice recognition system, such as sampling frequency, quantization accuracy, number of channels, etc., to improve the performance of the voice recognition chip.

[0068] After confirming that the call device has received the call request, a voice session will be established, so that the call device can detect the establishment of a voice session. Then the voice acquisition function of the voice recognition chip is started, and the voice recognition chip begins to receive voice signals from the audio input end of the call device and performs preliminary preprocessing, such as denoising and filtering. The voice recognition chip converts the preprocessed voice signal into a digital signal, which is transmitted to the voice gateway through the internal bus or external interface for further processing.

[0069] In addition, the incoming call number can be identified to determine whether it is within the preset range of numbers allowed for collection. If it is a recorded relative number, no voice collection will be performed. On the contrary, if it is an unfamiliar number, voice collection will be performed on the call.

[0070] Step S20: If a preset keyword is recognized in the voice data, a recording file corresponding to the voice conversation is generated.

[0071] In this embodiment, the voice recognition chip first receives the collected voice data from the audio input terminal of the call device, and then extracts the keywords in the voice data through the voice recognition chip built into the call device, wherein the voice data is the voice data generated by the user based on the call device and the initiator of the call request. The voice data is collected by the voice recognition chip, and the keywords in the voice data can be extracted based on the voice recognition chip.

[0072] Optionally, in this embodiment, if a preset keyword is recognized in the voice data, the step of generating a recording file corresponding to the voice conversation includes:

[0073] The collected voice data is subjected to noise reduction and enhancement processing according to the signal processing module to obtain target voice data; sentence endpoints in the target voice data are detected based on voice activation detection technology, and corresponding voice features are extracted; forward reasoning is performed based on the sentence endpoints and the voice features to obtain a posterior probability distribution, and keywords in the voice data are determined based on the posterior probability distribution; when the keyword matches a preset keyword, a recording file corresponding to the voice conversation is generated.

[0074] Specifically, the built-in speech recognition chip of the communication device is based on a deep neural network algorithm. First, the speech data is picked up from the microphone of the communication device, and then the signal processing module performs noise reduction, enhancement and sentence endpoint detection based on VAD (Voice Activity Detection) technology. Then the processed speech features are extracted, and the feature data is forward-inferred to obtain the posterior probability distribution of the phoneme. Finally, the keyword is determined based on the posterior probability distribution. The collected speech data is processed by using a digital filter.

[0075] The signal processing module can reduce noise and enhance the voice data. The collected voice data can be filtered by a digital filter to remove background noise. Specifically, the voice signal is first converted to the frequency domain using the fast Fourier transform (FFT), and then filtered in the frequency domain, and then converted back to the time domain through the inverse FFT, thereby completing the noise reduction of the voice data. The filtered voice data is gain controlled to improve the clarity and recognition of the voice signal. Specifically, an adaptive gain control algorithm can be used to dynamically adjust the gain according to the real-time changes of the voice signal. After noise reduction and enhancement, clear and highly recognizable target voice data is obtained.

[0076] The specific method of detecting the sentence endpoints in the target voice data based on the voice activation detection technology is to first process the target voice data using the VAD algorithm to detect the active segment (i.e., the segment containing voice information) and the inactive segment (i.e., the silent or background noise segment) in the voice signal. In the active segment, the starting point and the ending point of the sentence are further detected, and the sentence endpoints can be detected by using algorithms such as the energy threshold method and the zero-crossing rate method. Then, according to the detected sentence endpoints, the corresponding voice features are extracted from the target voice data, where the voice features include frequency, energy, zero-crossing rate, Mel-frequency cepstral coefficient (MFCC), etc.

[0077] Finally, forward reasoning is performed based on the sentence endpoints and speech features to build a speech model. The trained speech model, which can be a convolutional neural network CNN or a recurrent neural network RNN, is used to input the extracted speech features into the trained speech model for forward reasoning. The model will output the corresponding posterior probability distribution based on the input speech features. Among them, the posterior probability distribution represents the model's confidence in the input speech features corresponding to different text sequences. The output of the model can then be converted into a probability distribution form through the softmax function. The text sequence with the highest probability is selected from the posterior probability distribution as the recognition result. In the recognized text sequence, the target keyword is determined according to the preset keyword library or rules. In addition, the speech model needs to be trained with a large amount of speech data in advance to learn the mapping relationship between speech signals and text.

[0078] As an optional implementation method, the extracted feature vectors can be matched with the acoustic model to calculate the score of each feature vector in the acoustic feature. Based on the scoring results, the acoustic feature sequence corresponding to the most likely speech signal is determined. Then the probability of the sound signal corresponding to the phrase sequence is calculated based on the language model. These models usually take into account information such as the co-occurrence probability and grammatical structure of vocabulary. Phrase sequence decoding: The phrase sequence is decoded according to the acoustic feature sequence and the language model. A dynamic programming algorithm, such as the Viterbi algorithm, can be used. Finally, in the decoded phrase sequence, keywords are extracted according to a preset keyword extraction algorithm.

[0079] Further, in this embodiment, when the keyword matches the preset keyword, the step of generating a recording file corresponding to the voice conversation includes:

[0080] According to the keyword, a recognition result is determined in a decoding search space; according to the keyword, an accuracy rate of the recognition result is determined, and when the accuracy rate is higher than a preset threshold, it is determined that the keyword matches a preset keyword; and recording is started and the voice conversation is recorded to generate the recording file.

[0081] Specifically, after the keywords are extracted, the extracted keywords are matched with a preset keyword library. The preset keyword library is pre-defined and contains specific keywords or phrases that need to trigger the recording function. If the extracted keywords successfully match a keyword in the preset keyword library, the system proceeds to the next step; if the match fails, no recording is performed and the call continues normally.

[0082] Specifically, when a preset keyword matching the extracted keyword is found in the preset keyword library, after the match is successful, the telephone identification system needs to initialize the recording module, including configuring recording parameters (such as sampling rate, bit rate, etc.), and ensuring that the recording module is in normal working state. The recording module can be deployed in a voice gateway, which is in communication connection with the call device, and can record the call through the recording module.

[0083] In addition, a new recording file needs to be created for the call request to be recorded. This file is usually saved in a specified storage location and given a unique file name for subsequent management and search. After the recording device is initialized, the system starts recording the voice data corresponding to the call request. During the recording process, the system will continuously write voice data to the previously created recording file. After the recording is completed, the system will save the recording file to the specified storage location. This storage location can be a local hard disk, network storage, or other type of storage device. In order to facilitate subsequent search and management, the system usually names and classifies the recording files. Naming can be based on information such as recording time, call requester, keywords, etc.; classification can be based on factors such as call type and importance. In order to prevent the loss or damage of the recording file, the system can also back up the recording file regularly. The backup can be stored in another physical location or transferred to a remote server for storage over the network.

[0084] It should be noted that before recording a voice file, it is also necessary to request recording permission from the user corresponding to the call device. The corresponding request can be fed back to the front end of the call device and displayed in the front end interface. The user agrees to record the voice file based on the input device of the call device. After detecting the user's authorization, the call device will perform the action of recording the voice file.

[0085] Optionally, when determining whether there is a match with a preset keyword, a decoder is first designed, which can convert the input audio or text data into a series of keyword combinations or keyword sequences, wherein the decoder can be constructed based on an acoustic model or a language model. Then the extracted keywords are obtained and used as the basis of the search space. The input preset keyword library is decoded using the decoder to generate a search space containing all keyword combinations. Then at least one evaluation criterion is determined, which includes accuracy and recall.

[0086] By applying search algorithms such as depth-first search and breadth-first search in the search space, the optimal keyword recognition result is searched according to the set evaluation criteria. Then each recognition result obtained in the search process is evaluated, that is, according to actual needs, a preset threshold is set, and the preset threshold is saved in the evaluation criteria. The preset threshold is used to determine whether the recognition result meets the requirements. Specifically, the recognition result can be compared with the preset threshold. If the recognition result meets the threshold requirement, such as the accuracy is higher than a certain value, the recall rate is within a certain range, etc., it is regarded as the final result, and the keyword is determined to match the preset keyword, and then the step of starting recording and recording the recording file corresponding to the call request is executed.

[0087] In addition, the final result can also be output to the user interface of the telephone recognition system for further processing by the user. The user can view the output matching keywords and preset keywords in the user interface, and then feedback corresponding modification information based on the keyword matching situation in the user interface.

[0088] Step S30: Upload the recording file to the cloud server through the voice gateway, and receive the recognition result fed back by the cloud server.

[0089] In this embodiment, when the keyword matches the preset keyword, the voice gateway obtains the recorded audio file and then uploads the audio file to the cloud server. After the cloud server receives the audio file uploaded by the voice gateway, it recognizes the audio file through the large model deployed on the cloud server and feeds back the recognition result to the voice gateway.

[0090] Specifically, after completing the call recording, the telephone identification system will generate at least one recording file, which is usually stored in common audio formats such as MP3, WAV, etc. Before uploading, the integrity and readability of the recording file must be checked, as well as whether the file is damaged and the format is correct. During the upload process, the voice gateway will use the network connection to send the recording file to the specified cloud server address. The voice gateway serves as a bridge connecting the local device and the cloud server and is responsible for transmitting the recording file to the cloud.

[0091] After the cloud server receives the recording file uploaded by the voice gateway, it will perform file integrity verification and format check. If the file is damaged or the format is incorrect, the cloud server will return an error message. The cloud server uses the deployed speech recognition technology to recognize the recording file and generate the corresponding text recognition result. After the recognition is completed, the cloud server will return the recognition result to the voice gateway in text form.

[0092] Optionally, a large language model LLM (Large Language Model) is associated with the cloud server, and its principle is mainly based on deep learning and natural language processing technology. The training process of LLM mainly includes two stages: unsupervised pre-training and supervised fine-tuning. Among them, unsupervised pre-training is to train the model with a large amount of unlabeled text data so that it can capture the grammar, semantics and contextual relationships of natural language. Supervised fine-tuning is to fine-tune the model for specific tasks (such as text classification, named entity recognition, etc.) after completing pre-training. This stage usually requires adding some task-specific layers or parameters on the basis of the pre-trained model, and using labeled data for training.

[0093] As an optional implementation method for determining the recognition result of a recording file through LLM, first deploy the LLM server on the cloud server and ensure that the LLM server is in normal operation, that is, the pre-trained language model suitable for semantic and intent recognition has been loaded. After receiving the recording file, it is parsed and the audio data is converted into text data, which can be converted through automatic speech recognition technology. Then use LLM to perform semantic and intent recognition on the converted text data. LLM can analyze key information in the conversation content, such as requests, threats, promises, etc., by understanding the subtle complexity of context, meaning and language, so as to judge the intention of the call, and then determine the intent recognition result and semantic recognition result.

[0094] Step S40: Execute a processing action corresponding to the recognition result.

[0095] In this embodiment, after receiving the recognition result fed back by the cloud server, it is determined whether the phone corresponding to the call request is an abnormal phone through recognition. When the phone recognition system determines that the call request of the recognition result is an abnormal phone according to the preset processing logic, the corresponding processing action is performed on the call request.

[0096] It should be noted that before obtaining the telephone number of the caller, it is necessary to request permission from the user to obtain the telephone number. Similarly, the acquisition request can be fed back to the front-end interface corresponding to the call device, and the user operates the request based on the front-end interface. After the call device detects the user's authorization, it obtains the telephone number of the caller. The above user can be a guardian, and the corresponding call device is the guardian's terminal.

[0097] Specifically, according to the recognition results, search for keywords or phrases related to abnormal behavior. Compare with the preset abnormal call feature library to further confirm whether it is an abnormal call. After the judgment is completed by the phone recognition system, the judgment result can be sent to the review end for a second judgment to ensure the accuracy of the judgment result. The review end can be a manual review. Manual review can be performed by listening to recording files, checking call records, analyzing user feedback, etc.

[0098] Furthermore, in this embodiment, after step S40, the following steps are further included:

[0099] When receiving a misjudgment instruction fed back by a user, determining a misjudgment phone number corresponding to the misjudgment instruction; querying a blacklist library for a target number matching the misjudgment phone number and associated target voiceprint data; and deleting the target number and the target voiceprint data from the blacklist library.

[0100] Specifically, a user feedback channel is first established, and a feedback module is set in the phone identification system. The feedback module is associated with a corresponding feedback interface, and the user can feedback corresponding information to the phone identification system through the feedback interface. When the phone identification system receives the misjudgment instruction fed back by the user, the corresponding misjudged phone record is retrieved from the system log or database according to the misjudgment instruction information provided by the user to confirm the specific number of the misjudged phone. Then access the blacklist library, which should contain known bad or misjudged phone numbers and their associated information target numbers and target voiceprint data. Perform a query operation in the blacklist library, use the misjudged phone as the query keyword, and find the matching target number and the associated target voiceprint data. After confirmation, perform a deletion operation to remove the target number and its associated target voiceprint data from the blacklist library.

[0101] In addition, relevant information about the deletion operation, such as the deletion time, operator, and deletion reason, can be recorded for subsequent audit or inquiry. The configuration or status of the relevant system is then automatically updated to ensure that the misjudged call is no longer incorrectly intercepted or processed. The user is then notified that the misjudged call has been successfully processed and deleted from the blacklist.

[0102] In the technical solution provided in this embodiment, if it is detected that a voice conversation is established, the voice data corresponding to the voice conversation is collected through the voice recognition chip, and then if a preset keyword is recognized in the voice data, a recording file corresponding to the voice conversation is generated, and then the recording file is uploaded to the cloud server through the voice gateway, and the recognition result fed back by the cloud server is received, and finally the processing action corresponding to the recognition result is executed. This application collects voice data through a voice recognition chip for preliminary processing and judgment, and recognizes the recording file of the voice conversation in combination with the cloud server, thereby improving the accuracy of abnormal call recognition.

[0103] In the technical solution provided in this embodiment, by adding a low-cost local voice recognition chip to the telephone identification system, suspicious information can be identified through sensitive words before uploading the recording file to the cloud for further semantic and intent recognition, thereby reducing the cost of abnormal telephone identification.

[0104] Embodiment 2

[0105] Please refer to Figure 2 In the second embodiment, when the telephone number corresponding to the call request is determined to be an abnormal telephone number according to the recognition result, the step of executing the corresponding processing action includes:

[0106] Step S50: Obtaining semantic recognition results and intention recognition results in the recognition results.

[0107] Step S60: Determine whether the semantic recognition result and the intention recognition result match according to a preset abnormal phone feature library.

[0108] Step S70: When the semantic recognition result and the intention recognition result match the features in the abnormal phone feature library, the phone corresponding to the recognition result is determined to be an abnormal phone.

[0109] In this embodiment, the recognition result is fed back by the cloud server, which determines the recognition result of the recording file based on the deployed LLM model and feeds it back to the phone recognition system. The specific recognition process is shown in Example 1 and will not be repeated here. After the phone recognition system receives the recognition result, it obtains the semantic recognition result and the intention recognition result. Then, using the semantic recognition result and the intention recognition result as keywords, it queries in the preset abnormal phone feature library to see if there is a matching abnormal phone feature. If so, the phone corresponding to the recognition result is determined to be an abnormal phone.

[0110] Specifically, the recognition results are received and parsed from the speech recognition system to ensure that the recognition results contain two parts: semantic recognition results and intention recognition results. Among them, the semantic recognition result is the content after the speech is converted into text, reflecting the specific language information in the call, and the intention recognition result is the analysis of the user's intention in the speech content, such as requesting services, asking for information, and conducting transactions. Then obtain the abnormal call feature library, which should contain the semantic features and intention features of various abnormal calls. Semantic features can include common abnormal words, keywords, phrases, etc., and intention features include intentions to obtain personal information, request transfers, false service promises, etc. Then compare the semantic recognition results with the semantic features in the abnormal call feature library, which can be achieved through text matching algorithms (such as keyword matching, regular expression matching, etc.). At the same time, match the intention recognition results with the intention features in the feature library, and use a pre-trained model for matching.

[0111] During the matching process, different matching weights can be set. For example, for highly suspicious semantic and intent features, a lower threshold can be set to trigger an alarm. If both the semantic recognition result and the intent recognition result match the features in the abnormal phone feature library, that is, the preset matching weight is reached, the phone corresponding to the recognition result is determined to be an abnormal phone. If the matching result is not a complete match or does not meet the preset matching criteria, a further manual review step can be used to make a final judgment.

[0112] Step S80: Execute the processing action corresponding to the abnormal call result.

[0113] In this embodiment, when it is determined that the phone number corresponding to the recognition result is an abnormal phone number, the corresponding instruction is triggered, thereby executing the preset action associated with the instruction.

[0114] Optionally, in this embodiment, step S80 includes:

[0115] Receive a call termination instruction sent by the voice gateway to terminate the call process corresponding to the voice conversation; or, receive a prompt instruction sent by the voice gateway, and start a voice prompt function according to the prompt instruction; or, obtain an associated target phone through the voice gateway, and send a prompt voice to the target phone through a call; identify the phone number corresponding to the voice conversation, and determine the voiceprint data in the recording file, and associate the phone number and the voiceprint data and save them in a blacklist library.

[0116] Specifically, when it is determined that the phone corresponding to the call request is an abnormal phone, a call termination instruction can be sent to the call device through the voice gateway to cause the call device to terminate the call process corresponding to the call request, or a prompt instruction can be sent to the call device through the voice gateway to cause the call device to start the voice prompt function according to the instruction and remind the user who is talking that the phone is an abnormal call. Alternatively, the target phone associated with the call device can be obtained through the voice gateway, and the prompt voice can be sent to the target phone through the call, wherein the associated target phone here can be the phone number of the user's family. By identifying the phone number corresponding to the call request and determining the voiceprint data in the recording file, the phone number and the voiceprint data are associated and saved in the blacklist library.

[0117] In the technical solution provided in this embodiment, by obtaining the semantic recognition result and the intention recognition result in the recognition result, and then judging whether the semantic recognition result and the intention recognition result match according to the preset abnormal phone feature library, finally when the semantic recognition result and the intention recognition result match the features in the abnormal phone feature library, the phone corresponding to the recognition result is judged to be an abnormal phone, and the corresponding processing action is performed. This embodiment determines whether it matches the features in the abnormal phone feature library by receiving the recognition result from the cloud server, thereby judging whether it is an abnormal phone, so as to improve the accuracy of abnormal phone recognition.

[0118] Embodiment 3

[0119] Please refer to Figure 3 In a third embodiment, the method further includes:

[0120] Step S90: Identify the voiceprint of the recording file through the voice gateway, and compare the voiceprint with a pre-stored voiceprint blacklist library.

[0121] Step S100: When there is a target voiceprint matching the voiceprint in the voiceprint blacklist library, it is determined that the target phone corresponding to the voiceprint is an abnormal phone.

[0122] Step S110: Execute the processing action corresponding to the abnormal call result.

[0123] In this embodiment, the processing action is performed by the voice gateway. After the voiceprint of the identified abnormal phone is stored in the voiceprint blacklist library, it is determined whether the phone of the call request is an abnormal phone. The voiceprint of the recording file can be first identified, and the voiceprint is compared with the pre-stored voiceprint blacklist library. It is determined whether there is a target voiceprint matching the voiceprint in the voiceprint blacklist library. If so, it is determined that the target phone corresponding to the voiceprint is an abnormal phone, and the processing action corresponding to the abnormal phone is executed. The processing action here refers to the content in the second embodiment, and will not be repeated in detail.

[0124] Specifically, extract the voice segment to be identified from the recording file, and extract the sound features that can represent the individual's identity from the voice segment, such as Mel-frequency cepstral coefficients and perceptual linear prediction. Use machine learning algorithms to train and optimize the extracted sound features to obtain voiceprints. Compare the voiceprints with the voiceprint blacklist library to find the same target voiceprints.

[0125] In the technical solution provided by this embodiment, by identifying the voiceprint of the recording file and comparing the voiceprint with the pre-stored voiceprint blacklist library, when the target voiceprint matching the voiceprint exists in the voiceprint blacklist library, the target phone corresponding to the voiceprint is determined to be an abnormal phone, and the corresponding processing action is performed. When the recorded recording file is received again, it is directly determined whether it is an abnormal phone through the local voiceprint blacklist library, thereby improving the efficiency of abnormal phone identification.

[0126] Embodiment 4

[0127] Please refer to Figure 4 In the fourth embodiment, after the step of uploading the recording file to the cloud server through the voice gateway and receiving the recognition result fed back by the cloud server, it also includes:

[0128] Step S120: inputting the recording file into a voice imitation recognition model, and determining whether the recording file is AI synthesized speech according to the voice imitation recognition model.

[0129] Step S130: extracting timbre parameters of the recording file according to the voice imitation recognition model, and comparing the timbre parameters with pre-stored target timbre parameters.

[0130] Step S140: When the timbre parameters match the target timbre parameters, the recording file is determined to be AI synthesized speech, and a corresponding voice prompt is output.

[0131] In this embodiment, first ensure that the format of the recording file is compatible with the imitation voice recognition model. If not, the format of the recording file needs to be converted to the format required by the model. Load the trained imitation voice recognition model, and input the pre-processed recording file into the imitation voice recognition model. The model automatically extracts timbre parameters from the recording file, wherein the timbre parameters include sound quality, pitch, and sound intensity. The model compares the extracted timbre parameters with the pre-stored target timbre parameters to determine whether the recording file is a synthesized speech, wherein the target timbre parameters are pre-extracted from human voices. The imitation voice recognition model compares the timbre parameters with the pre-stored target timbre parameters, and when the timbre parameters match the target timbre parameters, it determines that the recording file is an AI synthesized speech. If the judgment result is a synthesized speech, prepare corresponding voice prompt content. For example, "This recording is a synthesized speech, please handle it with caution." Use speech synthesis technology to convert the prepared voice prompt content into a voice signal. The generated voice prompt signal is transmitted to the call device through a communication interface (such as a telephone line, a network interface, etc.). After receiving the voice prompt signal, the call device automatically plays it for the user to listen.

[0132] In the technical solution provided by this embodiment, the recording file is input into the voice imitation recognition model, and the voice imitation recognition model is used to determine whether the recording file is a synthesized voice, and then the judgment result output by the voice imitation recognition model is obtained. If the judgment result is a synthesized voice, a corresponding voice prompt is fed back to the communication device. This embodiment can determine whether the sound is made by a robot through the voice imitation recognition model, thereby improving the accuracy of abnormal call recognition.

[0133] Since the system introduced in the embodiment of the present application is a system used to implement the method of the embodiment of the present application, based on the method introduced in the embodiment of the present application, the person skilled in the art can understand the specific structure and deformation of the system, so it is not repeated here. All systems used in the method of the embodiment of the present application belong to the scope of protection of this application.

[0134] The present application provides a communication device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the control method of the communication device in the above-mentioned embodiment 1.

[0135] Reference below Figure 5, which shows a schematic diagram of the structure of a call device suitable for implementing the embodiment of the present application. The call device in the embodiment of the present application may include but is not limited to mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions: tablet computers), PMPs (Portable Media Players: portable multimedia players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 5 The communication device shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0136] like Figure 5 As shown, the communication device may include a processing device 1001 (e.g., a core processor, a graphics processor, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM: Read Only Memory) 1002 or a program loaded from a storage device 1003 to a random access memory (RAM: Random Access Memory) 1004. In RAM1004, various programs and data required for the operation of the communication device are also stored. The processing device 1001, ROM1002, and RAM1004 are connected to each other through a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Generally, the following systems can be connected to the I / O interface 1006: an input device 1007 including, for example, a touch screen, a touch pad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; an output device 1008 including, for example, a liquid crystal display (LCD: Liquid Crystal Display), a speaker, a vibrator, etc.; a storage device 1003 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 1009. The communication device 1009 can allow the communication device to communicate with other devices wirelessly or by wire to exchange data. Although the figure shows a communication device with various systems, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems can be implemented or have alternatively.

[0137] In particular, according to the embodiments disclosed in the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.

[0138] The communication device provided by the present application adopts the control method of the communication device in the above embodiment, which can solve the technical problem that the accuracy of abnormal phone recognition in the related solution is low. Compared with the prior art, the beneficial effects of the communication device provided by the present application are the same as the beneficial effects of the control method of the communication device provided by the above embodiment, and the other technical features in the communication device are the same as the features disclosed in the method of the previous embodiment, which will not be repeated here.

[0139] It should be understood that the various parts disclosed in this application can be implemented by hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0140] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0141] The present application provides a computer-readable storage medium having computer-readable program instructions (ie, computer programs) stored thereon, and the computer-readable program instructions are used to execute the control method of the communication device in the above-mentioned embodiment.

[0142] The computer-readable storage medium provided in the present application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared or semiconductor systems, systems or devices, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM: Random Access Memory), a read-only memory (ROM: Read Only Memory), an erasable programmable read-only memory (EPROM: Erasable Programmable Read Only Memory or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM: CD-Read Only Memory), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency: Radio Frequency), etc., or any suitable combination of the above.

[0143] The computer-readable storage medium may be included in the communication device; or may exist independently without being installed in the communication device.

[0144] The above-mentioned computer-readable storage medium carries one or more programs. When the above-mentioned one or more programs are executed by the communication device, the communication device: if it detects that a voice conversation is established, collects voice data corresponding to the voice conversation through the voice recognition chip; if preset keywords are recognized in the voice data, generates a recording file corresponding to the voice conversation; uploads the recording file to the cloud server through the voice gateway, and receives the recognition result fed back by the cloud server; and executes the processing action corresponding to the recognition result.

[0145] Computer program code for performing the operations of the present application may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0146] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of a code, and the module, a program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0147] The modules involved in the embodiments described in this application may be implemented by software or hardware, wherein the name of the module does not constitute a limitation on the unit itself in some cases.

[0148] The readable storage medium provided by the present application is a computer-readable storage medium, which stores computer-readable program instructions (i.e., computer programs) for executing the control method of the above-mentioned communication device, and can solve the technical problem that the accuracy of abnormal phone identification in the related scheme is low. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided by the present application are the same as the beneficial effects of the control method of the communication device provided by the above-mentioned embodiment, and will not be repeated here.

[0149] An embodiment of the present application provides a computer program product, including a computer program, which implements the steps of the control method of the communication device as described above when the computer program is executed by a processor.

[0150] The computer program product provided by the present application can solve the technical problem that the accuracy of abnormal phone identification in related solutions is low. Compared with the prior art, the beneficial effects of the computer program product provided by the embodiment of the present application are the same as the beneficial effects of the control method of the call device provided by the above embodiment, which will not be repeated here.

[0151] The above are only preferred embodiments of the present application, and are not intended to limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent processing scope of the present application.

Claims

1. A method for controlling a communication device, characterized in that: Applied to a communication device, the communication device is provided with a voice recognition chip and is connected to a cloud server through a voice gateway. The control method of the communication device includes the following steps: If it is detected that a voice conversation is established, voice data corresponding to the voice conversation is collected by the voice recognition chip; If a keyword matching a preset keyword is identified in the voice data, determining a recognition result in a decoding search space according to the keyword; Determining the accuracy of the recognition result according to the keyword, and when the accuracy is higher than a preset threshold, determining that the keyword matches a preset keyword; Starting recording and recording the voice conversation to generate a recording file; Uploading the recording file to the cloud server through the voice gateway, and receiving the recognition result fed back by the cloud server, wherein the recognition result includes a semantic recognition result and an intent recognition result; Obtaining a semantic recognition result and an intention recognition result from the recognition result; Determining whether the semantic recognition result and the intention recognition result match according to a preset abnormal phone feature library; When the semantic recognition result and the intention recognition result match the features in the abnormal phone feature library, determining that the phone corresponding to the recognition result is an abnormal phone; Execute the processing action corresponding to the abnormal call identified as a result; Identify the phone number corresponding to the voice conversation, determine the voiceprint data in the recording file, and associate the phone number and the voiceprint data and store them in a blacklist database; Inputting the recording file into a voice imitation recognition model, and determining whether the recording file is AI synthesized speech according to the voice imitation recognition model; Extracting the timbre parameters of the recording file according to the sound imitation recognition model, and comparing the timbre parameters with pre-stored target timbre parameters; When the timbre parameters match the target timbre parameters, the recording file is determined to be AI synthesized speech, and a corresponding voice prompt is output.

2. The method according to claim 1, characterized in that The step of identifying a keyword matching a preset keyword in the voice data comprises: The speech recognition chip performs noise reduction and enhancement processing on the collected speech data according to the signal processing module to obtain target speech data; Detecting sentence endpoints in the target voice data based on voice activation detection technology and extracting corresponding voice features; Forward reasoning is performed based on the sentence endpoints and the speech features to obtain a posterior probability distribution, and keywords in the speech data are determined based on the posterior probability distribution.

3. The method according to claim 1, characterized in that The step of executing the processing action corresponding to the abnormal call as the identification result comprises: Receiving a call termination instruction sent by the voice gateway to terminate the call process corresponding to the voice session; Alternatively, receiving a prompt instruction sent by the voice gateway, and starting a voice prompt function according to the prompt instruction; Alternatively, the associated target phone is obtained through the voice gateway, and the prompt voice is sent to the target phone through the call.

4. The method according to claim 1, characterized in that The method further comprises: Identifying the voiceprint of the recording file through the voice gateway, and comparing the voiceprint with a pre-stored voiceprint blacklist library; When there is a target voiceprint matching the voiceprint in the voiceprint blacklist library, determining that the target phone corresponding to the voiceprint is an abnormal phone; Execute the corresponding processing action for abnormal calls.

5. The method according to claim 1, characterized in that After the step of executing the processing action corresponding to the abnormal call as the identification result, the method further includes: Upon receiving a misjudged instruction fed back by a user, determining a misjudged phone number corresponding to the misjudged instruction; Searching the blacklist database for the target number that matches the misjudged phone number and the associated target voiceprint data; The target number and the target voiceprint data are deleted from the blacklist database.

6. A communication device, characterized in that: The communication device comprises: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the control method of the communication device according to any one of claims 1 to 5.

7. A storage medium, characterized in that: The storage medium is a computer-readable storage medium, and a computer program is stored on the computer-readable storage medium. When the computer program is executed by a processor, the steps of the control method of the communication device according to any one of claims 1 to 5 are implemented.

Images