Method, device, control panel, and equipment for filtering sensitive words in speech recognition input

Abstract

The present invention relates to a sensitive word filtering method, device, control panel, and equipment for voice recognition input, and belongs to the field of information technology of the intelligent device industry. The method comprises: performing voice recognition processing on the input voice information to obtain the voice recognition result corresponding to the voice information, and simultaneously performing sensitive word monitoring processing on the input voice information using a sensitive word monitoring circuit to obtain the sensitive word monitoring result corresponding to the voice information; performing sensitive word filtering on the voice recognition result according to the sensitive word monitoring result to obtain the voice recognition result after sensitive word filtering. The present invention realizes sensitive word filtering based on the sensitive word monitoring circuit, which can effectively reduce the consumption of software and hardware resources of the main function of voice recognition by sensitive word filtering, is conducive to the normal realization of the main function of voice recognition, and reduces the influence of network delay on the sensitive word screening effect, thereby enhancing the reliability and stability of sensitive word filtering.

Description

Background Art

[0002] As self-service equipment in industries such as finance, telecommunications, and healthcare becomes increasingly intelligent, voice recognition input has become the primary mode of human-computer interaction in AI systems.

[0003] The current method for filtering sensitive words in voice recognition input mainly relies on the software and hardware resources of the main function of voice recognition input to implement the sensitive word screening and deletion functions. For example, the input voice information is firstly subjected to voice recognition to obtain the voice recognition result, and then the voice recognition result is filtered and deleted according to the preset sensitive word list to obtain the voice recognition result after sensitive word filtering.

[0004] However, the above-mentioned method for filtering sensitive words for speech recognition input mainly relies on the software and hardware resources of speech recognition, and the network delay affects the main function of speech recognition. Therefore, the screening effect of sensitive words is unstable, which is not conducive to the normal implementation of the main function of speech recognition input. Summary of the invention

[0005] In order to solve the defects of the existing method for filtering sensitive words for voice recognition input, that is, the screening effect of sensitive words is unstable and is not conducive to the normal implementation of the main function of voice recognition input, the present invention provides a method, device, control panel and equipment for filtering sensitive words for voice recognition input.

[0006] In a first aspect, in order to solve the above technical problem, the present invention provides a method for filtering sensitive words for speech recognition input, comprising:

[0007] Performing voice recognition processing on the input voice information to obtain a voice recognition result corresponding to the voice information, and simultaneously performing sensitive word monitoring processing on the input voice information using a sensitive word monitoring circuit to obtain a sensitive word monitoring result corresponding to the voice information;

[0008] According to the sensitive word monitoring result, the speech recognition result is filtered for sensitive words to obtain a speech recognition result after sensitive word filtering.

[0009] The beneficial effect of the present invention is that: while performing voice recognition processing on the input voice information, an independent sensitive word monitoring circuit is used to implement the sensitive word monitoring function, so that sensitive words are screened out from the input voice information, which can effectively reduce the consumption of software and hardware resources of the main function of voice recognition by sensitive word filtering, is conducive to the normal implementation of the main function of voice recognition, and reduces the influence of network delay on the sensitive word screening effect, thereby enhancing the reliability and stability of sensitive word filtering.

[0010] Furthermore, the process of obtaining the sensitive word monitoring result includes:

[0011] The voice recognition chip in the sensitive word monitoring circuit is used to perform sensitive word monitoring processing on the input voice information to obtain a sensitive word monitoring result corresponding to the voice information.

[0012] The beneficial effect of adopting the above-mentioned improvement scheme is: the hardware implementation of voice monitoring and recognition is completed by using the sensitive word monitoring circuit of the above-mentioned structure, that is, the hardware implementation of the sensitive word monitoring function. Compared with the soft implementation that relies on the voice recognition software development toolkit and the network, it can avoid the consumption of software and hardware resources of the main function of voice recognition, and the screening performance of sensitive words will not involve the operating system performance and network speed, which can effectively improve the real-time performance of the monitoring function, and thus improve the sensitive word filtering performance.

[0013] Further, the using of the voice recognition chip in the sensitive word monitoring circuit to perform sensitive word monitoring processing on the input voice information to obtain a sensitive word monitoring result corresponding to the voice information includes:

[0014] A voice monitoring microphone is used to obtain first voice information from a voice input source, and based on a preset sensitive word list, the voice recognition chip is used to obtain sensitive word information from the first voice information, and the sensitive word monitoring result is generated according to the sensitive word information.

[0015] The beneficial effect of adopting the above-mentioned improvement scheme is: based on the preset sensitive word list, an independent voice recognition chip is used to perform sensitive word monitoring processing on the input voice information, so as to filter out the preset sensitive words from the input voice information, thereby flexibly setting the monitoring function for related sensitive words.

[0016] Further, the performing speech recognition processing on the input speech information to obtain a speech recognition result corresponding to the speech information includes:

[0017] The second voice information is acquired from the voice input source using a voice recognition main microphone, and the second voice information is converted into corresponding text data based on a voice recognition software development kit to obtain the voice recognition result.

[0018] The beneficial effect of adopting the above-mentioned improvement scheme is: using two independent microphones for voice recognition input and sensitive word monitoring input respectively can avoid the circuit adaptation problem caused by using a single microphone, thereby simplifying the circuit and facilitating implementation and application.

[0019] Further, the sensitive word monitoring result includes a sensitive word index, and the sensitive word filtering is performed on the speech recognition result according to the sensitive word monitoring result to obtain the speech recognition result after the sensitive word filtering, including:

[0020] The sensitive words corresponding to the sensitive word index are deleted from the speech recognition result to obtain the speech recognition result after the sensitive words are filtered.

[0021] The beneficial effect of adopting the above-mentioned improvement scheme is: using sensitive word indexes for data transmission can simplify data to improve data transmission efficiency, thereby further improving the real-time performance of sensitive word monitoring.

[0022] Further, the sensitive word monitoring circuit is used to perform sensitive word monitoring processing on the input voice information to obtain a sensitive word monitoring result corresponding to the voice information, and further includes:

[0023] According to the sensitive word information in the sensitive word monitoring result, a speech synthesis chip in the sensitive word monitoring circuit is used to generate sensitive word alarm voice information containing the sensitive word information, and a sensitive word alarm voice signal is issued based on the sensitive word alarm voice information.

[0024] The beneficial effect of adopting the above-mentioned improvement scheme is: based on an independent speech synthesis chip, the monitoring results are fed back to the user in real time without occupying normal speech recognition resources.

[0025] Furthermore, before filtering the speech recognition result for sensitive words according to the sensitive word monitoring result to obtain the speech recognition result after the sensitive word filtering, the method further includes:

[0026] The sensitive word monitoring result is sent to the application host.

[0027] The beneficial effect of adopting the above-mentioned improvement scheme is that the sensitive word monitoring circuit can provide correction information to the speech recognition main function module in real time, and feed back the sensitive word monitoring results to the application host in real time for alarm.

[0028] In a second aspect, the present invention provides a device for filtering sensitive words in speech recognition input, comprising a sensitive word monitoring module and a general speech recognition module;

[0029] The sensitive word monitoring module is used to call the sensitive word monitoring circuit to perform sensitive word monitoring processing on the input voice information to obtain the sensitive word monitoring result corresponding to the voice information;

[0030] The general speech recognition module is used to perform speech recognition processing on the speech information to obtain a speech recognition result corresponding to the speech information; and to perform sensitive word filtering on the speech recognition result according to the sensitive word monitoring result to obtain a speech recognition result after sensitive word filtering.

[0031] In a third aspect, the present invention provides a control panel for filtering sensitive words in voice recognition input, wherein the control panel is used to perform voice recognition processing on input voice information to obtain a voice recognition result corresponding to the voice information;

[0032] The control panel is also used to receive a sensitive word monitoring result after the voice information is monitored for sensitive words from the controller, and perform sensitive word filtering on the voice recognition result according to the sensitive word monitoring result to obtain a voice recognition result after sensitive word filtering.

[0033] In a fourth aspect, the present invention further provides a device for filtering sensitive words in speech recognition input, comprising the control panel as described above, and also comprising a controller;

[0034] The controller is used to perform sensitive word monitoring processing on the input voice information using the sensitive word monitoring circuit, obtain the sensitive word monitoring result corresponding to the voice information, and send the sensitive word monitoring result to the control board. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 A flowchart of a method for filtering sensitive words in speech recognition input provided by an embodiment of the present invention;

[0036] Figure 2 A schematic diagram of the workflow of thread 1 in a control panel provided by an embodiment of the present invention;

[0037] Figure 3 A schematic diagram of the workflow of thread 2 in the control board provided by an embodiment of the present invention;

[0038] Figure 4 A schematic diagram of the hardware structure of a method for implementing sensitive word filtering in speech recognition input provided by an embodiment of the present invention;

[0039] Figure 5 A schematic diagram of the working process of the MCU controller in the sensitive word monitoring circuit provided by an embodiment of the present invention;

[0040] Figure 6 A schematic diagram of the structure of an apparatus for filtering sensitive words in speech recognition input provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0041] The following examples are provided to further explain and supplement the present invention and do not constitute any limitation to the present invention.

[0042] A method for filtering sensitive words in speech recognition input according to an embodiment of the present invention is described below in conjunction with the accompanying drawings.

[0043] Reference Figure 1As shown, the present invention provides a method for filtering sensitive words for speech recognition input, comprising:

[0044] S1. Performing voice recognition processing on the input voice information to obtain a voice recognition result corresponding to the voice information, and simultaneously performing sensitive word monitoring processing on the input voice information using a sensitive word monitoring circuit to obtain a sensitive word monitoring result corresponding to the voice information;

[0045] S2. According to the sensitive word monitoring result, the speech recognition result is filtered for sensitive words to obtain a speech recognition result after sensitive word filtering.

[0046] Specifically, in this embodiment, the input voice information is mainly used to realize human-computer interaction based on voice recognition input. The voice information can be a voice request issued by the user, such as "turn on the air conditioner". The voice request can be converted into text data after voice recognition processing, thereby obtaining a voice recognition result.

[0047] Among them, while the voice request is converted into text data through voice recognition processing, the voice request is input into the sensitive word monitoring circuit in real time for sensitive word monitoring processing. If there are sensitive words in the voice request, such as swear words, dirty words and banned words, the sensitive words are filtered out from the voice request to obtain sensitive word monitoring results, and then the sensitive words are correspondingly eliminated from the voice recognition results according to the sensitive word monitoring results to obtain the voice recognition results after sensitive word filtering.

[0048] The above-mentioned embodiment provides a method for filtering sensitive words for voice recognition input. While performing voice recognition processing on the input voice information, an independent sensitive word monitoring circuit is used to implement the sensitive word monitoring function, so as to obtain sensitive words from the input voice information in real time. This can effectively reduce the consumption of software and hardware resources of the main function of voice recognition by sensitive word filtering, is conducive to the normal implementation of the main function of voice recognition, and reduces the impact of network delay on the sensitive word screening effect, thereby enhancing the reliability and stability of sensitive word filtering.

[0049] It is understandable that when the user completes the voice information input, all the voice recognition results after the sensitive words are filtered can be input into the application host to realize voice interaction. Therefore, the above-mentioned sensitive word filtering method of voice recognition input can be widely used in voice interaction scenarios of application hosts such as offline self-service equipment and service robots, and can effectively prevent sensitive words such as swear words, dirty words, and banned words from being input into the application system and causing related adverse effects.

[0050] Optionally, in one embodiment, the process of obtaining the sensitive word monitoring result includes:

[0051] The voice recognition chip in the sensitive word monitoring circuit is used to perform sensitive word monitoring processing on the input voice information to obtain a sensitive word monitoring result corresponding to the voice information.

[0052] The speech recognition chip may be a speech recognition integrated chip LD3320, so that the sensitive word monitoring circuit can complete the hardware implementation of speech monitoring and recognition through an independent speech recognition integrated chip, that is, the hardware implementation of sensitive word monitoring.

[0053] Compared with the soft implementation that relies on the speech recognition software development kit (SDK) and the network, the hard implementation has almost no recognition delay, and the recognition performance does not involve the operating system performance and network speed. Therefore, it can effectively improve the real-time performance of sensitive word monitoring, thereby avoiding the defects that affect the on-site user experience, such as system confusion, caused by the use of the same set of resources for speech recognition and sensitive word screening, and the influence of network delays when filtering sensitive words. This effectively improves the real-time performance of the monitoring function and then improves the sensitive word filtering performance.

[0054] Preferably, the using of the voice recognition chip in the sensitive word monitoring circuit to perform sensitive word monitoring processing on the input voice information to obtain a sensitive word monitoring result corresponding to the voice information includes:

[0055] A voice monitoring microphone is used to obtain first voice information from a voice input source, and based on a preset sensitive word list, the voice recognition chip is used to obtain sensitive word information from the first voice information, and the sensitive word monitoring result is generated according to the sensitive word information.

[0056] Wherein, the sensitive word monitoring circuit includes a voice recognition chip, and the preset sensitive word table includes at least one sensitive word. The preset sensitive word table can be pre-set in the voice recognition chip so as to filter out the preset sensitive words from the voice information.

[0057] Specifically, in this embodiment, the monitoring microphone is used to obtain first voice information from a voice input source. The voice input source may be a user who performs voice interaction with an application host. The monitoring microphone collects voice requests issued by the user to the application host in real time, and inputs the collected voice request content as the first voice information into the sound source pin of the voice recognition chip. The voice recognition chip then performs sensitive word screening on the first voice information to obtain at least one sensitive word. The at least one sensitive word is the sensitive word information. The sensitive word information may be subjected to processes such as sensitive word deduplication and sorting to obtain a sensitive word monitoring result.

[0058] In a possible implementation, the first voice information collected by the voice monitoring microphone can also be input into the voice recognition SDK, that is, the voice information collected by the same microphone is divided into two inputs, one input for voice recognition and the other input for sensitive word monitoring.

[0059] Among them, the voice recognition SDK can be installed using the control board, and a compatible adapter circuit is set between the voice monitoring microphone and the control board, so that the first voice information collected by the voice monitoring microphone is input into the audio input interface of the control board.

[0060] In a possible implementation, Figure 2 and Figure 3 As shown, the control board equipped with a voice recognition SDK can work in a dual-threaded manner. Thread 1 is used to call a third-party voice recognition SDK to complete the main function of voice recognition, and thread 2 is used to receive sensitive word alarm information issued by the sensitive word monitoring circuit, that is, the sensitive word monitoring result, and is responsible for modifying the voice recognition cache content, that is, filtering the cached voice recognition results for sensitive words.

[0061] For example, Figure 2 As shown, when the user performs voice interaction, thread 1 executes the following steps:

[0062] S101, thread 1 detects in real time whether there is voice input, if so, executes S102;

[0063] S102, calling a third-party voice SDK to perform voice recognition and convert it into text, and executing S103;

[0064] S103, check whether there is a speech end mark, if not, execute S104, if yes, execute S105;

[0065] S104, putting the obtained text into the cache, and returning to S101;

[0066] S105: Send the cached text to the application host.

[0067] It should be noted that when the voice interaction user makes voice input, it is actually the same voice input source, which is input into two recognition channels through two microphones respectively. One is the main voice recognition channel implemented by thread 1, and the other is the real-time sensitive word monitoring channel implemented by the sensitive word monitoring circuit.

[0068] Specifically, in this embodiment, thread 1 is responsible for listening for voice input and calling a third-party voice recognition SDK to convert voice recognition into text. For example, Baidu's voice SDK can be used to recognize and cache the user's real-time voice input sentence by sentence until the user completes the voice input, that is, when thread 1 detects the voice end mark, the cached text is transmitted to the application host through the communication interface (serial port, USB, network port, etc.) for application processing.

[0069] The speech end mark can be freely set according to actual needs. For example, the key "#0*" on the application host can be used as the speech end mark.

[0070] When thread 1 executes the above steps S101 to S105, Figure 3 As shown, thread 2 executes synchronously as follows:

[0071] S201, thread 2 detects whether there is sensitive word alarm information on the serial port, if so, executes step S202;

[0072] S202: Found sensitive words in the text cache and deleted them in real time.

[0073] It should be noted that when the input voice information contains preset sensitive words, the sensitive word monitoring circuit sends sensitive word alarm information, that is, the sensitive word monitoring result, to Thread 2 through the serial port. Therefore, Thread 2 can detect in real time whether there is a sensitive word alarm information on the serial port to know whether there is a sensitive word in the currently input voice information, so as to perform corresponding sensitive word filtering operations.

[0074] Specifically, in this embodiment, the sensitive word alarm information includes a sensitive word index. Thread 1 stores the voice recognition result in a text cache. At the same time, the voice recognition chip in the sensitive word monitoring circuit recognizes the sensitive word, generates a sensitive word index and transmits it to the controller in the sensitive word monitoring circuit. At this time, the controller can directly send the sensitive word index to Thread 2 in the control board, or it can first convert the sensitive word index into sensitive word alarm data including sensitive words and then send it to Thread 2. In addition, the controller can also alarm the application host separately through the serial port.

[0075] Among them, after thread 2 in the control board receives the sensitive word alarm information reported by the controller, it immediately finds the reported sensitive word in the text cache and deletes it from the cache, thereby ensuring that when thread 1 detects the voice end mark, the text cache sent to the application host does not contain the reported sensitive word.

[0076] Preferably, performing speech recognition processing on the input speech information to obtain a speech recognition result corresponding to the speech information includes:

[0077] The second voice information is acquired from the voice input source using a voice recognition main microphone, and the second voice information is converted into corresponding text data based on a voice recognition software development kit to obtain the voice recognition result.

[0078] Specifically, in this embodiment, two channels of collection and input are simultaneously performed for the voice information emitted by the same voice input source. One channel uses a third-party voice recognition SDK to realize voice recognition through a voice recognition main microphone, and the other channel uses a voice recognition chip in a sensitive word monitoring circuit to realize sensitive word monitoring through a voice monitoring microphone. By using two microphones for voice recognition input and sensitive word monitoring input respectively, the circuit adaptation problem caused by using a single microphone can be avoided, thereby simplifying the circuit and facilitating implementation and application.

[0079] Among them, since the two recognition channels of speech recognition and sensitive word monitoring are implemented independently, the sensitive words can be verified twice based on the recognition results of the two recognition channels, thereby ensuring the validity of the sensitive words screened out by the sensitive word monitoring circuit.

[0080] For example, the voice recognition channel of the sensitive word monitoring circuit and the voice recognition channel of the main voice recognition using the third-party SDK use different microphone inputs respectively, so the two voice recognition channels are completely independent. If one channel recognizes the sensitive word "AA", then the other channel should also recognize "AA" so that "AA" will be deleted.

[0081] If the main voice recognition channel does not recognize "AA", but the sensitive word monitoring circuit recognizes "AA", then "AA" may not be the sensitive word input by the same voice input source, and "AA" does not exist in the current voice recognition result. Therefore, "AA" will not enter the application host for processing, thereby ensuring the effectiveness of sensitive word filtering.

[0082] Optionally, in one embodiment, the sensitive word monitoring result includes a sensitive word index, and the sensitive word filtering is performed on the speech recognition result according to the sensitive word monitoring result to obtain the speech recognition result after sensitive word filtering, including:

[0083] The sensitive words corresponding to the sensitive word index are deleted from the speech recognition result to obtain the speech recognition result after the sensitive words are filtered.

[0084] It should be noted that in order to improve data transmission efficiency, when sensitive words that need to be monitored are preset, serial numbers can be assigned to the sensitive words, and sensitive words can be screened from voice information using a sensitive word monitoring circuit, and a sensitive word index can be generated based on the serial number corresponding to each sensitive word screened.

[0085] Specifically, in this embodiment, for each sensitive word screened from the voice information, the sensitive word monitoring circuit is used to look up the table to obtain the serial number corresponding to the sensitive word. For example, for sensitive word AA, the serial number of sensitive word AA obtained by looking up the table is 1, and the serial numbers of all sensitive words obtained by looking up the table are summarized to obtain a sensitive word index, and then the sensitive word index is sent from the sensitive word monitoring circuit to an execution module for deleting sensitive words from the voice recognition result. The execution module looks up the table based on the sensitive word index to obtain the corresponding sensitive word for sensitive word filtering.

[0086] Preferably, the method of performing sensitive word monitoring processing on the input voice information by using a sensitive word monitoring circuit to obtain a sensitive word monitoring result corresponding to the voice information further includes:

[0087] According to the sensitive word information in the sensitive word monitoring result, a speech synthesis chip in the sensitive word monitoring circuit is used to generate sensitive word alarm voice information containing the sensitive word information, and a sensitive word alarm voice signal is issued based on the sensitive word alarm voice information.

[0088] Specifically, in this embodiment, the speech synthesis chip reads the sensitive word information in the sensitive word monitoring result, and performs speech segment synthesis on the sensitive word information, thereby generating sensitive word alarm voice information containing the sensitive word information, such as the sensitive word alarm voice information of "Please do not voice input sensitive word AA", and then sends a sensitive word alarm voice signal to the user by playing the sensitive word alarm voice information by voice.

[0089] In one possible implementation, the sensitive word monitoring circuit can be connected to a speaker to play the sensitive word alarm voice information generated by the speech synthesis chip, so that the sensitive word monitoring circuit can independently implement sensitive word monitoring functions such as sensitive word screening and sensitive word alarm without occupying normal speech recognition and playback resources.

[0090] Preferably, the sensitive word filtering of the speech recognition result according to the sensitive word monitoring result to obtain the speech recognition result after the sensitive word filtering further comprises:

[0091] The sensitive word monitoring result is sent to the application host.

[0092] In this embodiment, by sending the sensitive word monitoring result to the application host for alarm, the risk of sensitive words entering the application system can be further reduced, and users are encouraged to use the self-service equipment in a civilized manner.

[0093] An embodiment of the present invention provides a control panel for filtering sensitive words in voice recognition input, wherein the control panel is used to perform voice recognition processing on input voice information to obtain a voice recognition result corresponding to the voice information;

[0094] The control panel is also used to receive a sensitive word monitoring result after the voice information is monitored for sensitive words from the controller, and perform sensitive word filtering on the voice recognition result according to the sensitive word monitoring result to obtain a voice recognition result after sensitive word filtering.

[0095] Optionally, the control board is also used to obtain second voice information from the voice input source using the voice recognition main microphone, and convert the second voice information into corresponding text data based on the voice recognition software development kit to obtain the voice recognition result.

[0096] Optionally, the sensitive word monitoring result includes a sensitive word index; the control panel is further used to delete the sensitive words corresponding to the sensitive word index from the speech recognition result to obtain the speech recognition result after the sensitive words are filtered.

[0097] An embodiment of the present invention provides a device for filtering sensitive words in voice recognition input, comprising the control panel described in the above embodiment, and also comprising a controller;

[0098] The controller is used to perform sensitive word monitoring processing on the input voice information using the sensitive word monitoring circuit, obtain the sensitive word monitoring result corresponding to the voice information, and send the sensitive word monitoring result to the control board.

[0099] Optionally, the controller is specifically used to utilize the voice recognition chip in the sensitive word monitoring circuit to perform sensitive word monitoring processing on the input voice information to obtain a sensitive word monitoring result corresponding to the voice information.

[0100] Optionally, the controller is also used to obtain first voice information from a voice input source using a voice monitoring microphone, obtain sensitive word information from the first voice information based on a preset sensitive word list using the voice recognition chip, and generate the sensitive word monitoring result according to the sensitive word information.

[0101] Optionally, the controller is further used to generate sensitive word alarm voice information containing the sensitive word information according to the sensitive word information in the sensitive word monitoring result by using the voice synthesis chip in the sensitive word monitoring circuit, and to issue a sensitive word alarm voice signal based on the sensitive word alarm voice information.

[0102] Optionally, the controller is further configured to send the sensitive word monitoring result to an application host.

[0103] For example, refer to Figure 4As shown, the above-mentioned method for filtering sensitive words for voice recognition input can be implemented by a hardware structure composed of a device for filtering sensitive words for voice recognition input and an application host.

[0104] The controller is a microcontroller unit (MCU) controller, the speech recognition chip uses the LD3320 speech recognition chip, and the speech synthesis chip uses the LQ-SYN6288TTS speech synthesis chip. The three components are as follows: Figure 4 The sensitive word monitoring circuit is shown in the dotted box.

[0105] It should be noted that the control board can be equipped with an operating system (such as Android system, Windows system or Linux system, etc.) and a third-party voice recognition SDK interface (such as Baidu, iFlytek, etc.) to realize the main function of voice recognition. In this embodiment, the control board is an Android control board, and the Android control board is provided with a network interface as a third-party voice recognition SDK interface.

[0106] Among them, Figure 4 As shown, dual voice input recognition microphones (a main sound receiving microphone and a sensitive word monitoring microphone) and related interfaces and communication lines are used to form a device for sensitive word filtering of voice recognition input.

[0107] Specifically, in this embodiment, an Android control board equipped with a third-party voice recognition SDK is used to implement the voice recognition function and transmit the result to the application host, and an MCU controller, an LD3320 voice recognition chip, and an LQ-SYN6288TTS voice synthesis chip are used to form a sensitive word monitoring circuit. Independent of the main radio microphone, an additional sensitive word monitoring microphone is used to monitor and screen the sensitive words in the input voice in real time, thereby forming a hardware structure of a dual-channel voice recognition channel with one working and one monitoring.

[0108] Among them, the MCU controller can also be a single-chip microcomputer. The LD3320 speech recognition chip and the LQ-SYN6288TTS speech synthesis chip are connected to the MCU controller via the SPI interface and control line, the serial port and control line respectively, so that the MCU controller can transmit and control data to the speech recognition chip and the speech synthesis chip.

[0109] For example, Figure 5 As shown, when the sensitive word monitoring circuit starts working, the MCU controller executes the following steps:

[0110] S301, MCU controller initialization, execution S302;

[0111] S302, set the LD3320 voice recognition chip to monitor the voice table, and execute S303;

[0112] S303, start polling to check alarm interruption, and execute S304;

[0113] S304, detecting whether an alarm interrupt is received, if so, executing S305;

[0114] S305, read the sensitive word index, send the sensitive word to the Android control board through the real-time alarm serial port 1, and execute S306;

[0115] S306, check whether the alarm application host is preset, if yes, execute S307, if not, execute S308;

[0116] S307, sending the sensitive words to the application host through the real-time alarm serial port 2, and executing S308;

[0117] S308. Play the alarm voice based on the LQ-SYN6288TTS voice synthesis chip, and return to S304.

[0118] Specifically, in this embodiment, the MCU controller is first initialized, and then a monitoring voice table containing preset sensitive words is set for the LD3320 voice recognition chip. The MCU controller starts to poll for alarm interrupts. If an alarm interrupt is received, that is, the preset sensitive words are found in the voice input, the LD3320 voice recognition chip sends a sensitive word index to the MCU controller.

[0119] The MCU controller reads the sensitive words and sends them to the Android controller through the alarm serial port 1, and decides whether to notify the application host according to the setting requirements. If there is a preset alarm application host, the MCU controller sends the sensitive words to the corresponding application host through the alarm serial port 2.

[0120] Among them, if the sensitive word monitoring circuit screens out the preset sensitive words, it will notify the control board in real time. The control board will remove the sensitive words from the current voice recognition cache results to ensure that when the voice recognition results are sent to the application host, the sensitive word content will not be input into the application system through the communication interface for processing.

[0121] The MCU controller on the sensitive word monitoring circuit will also control the LQ-SYN6288TTS speech synthesis chip to feed back sensitive word restriction information to the voice inputter through the voice speaker. If the sensitive word monitoring circuit does not filter out any sensitive words, when the voice input ends, the Android control board will send all current voice recognition results to the application host.

[0122] The sensitive word monitoring circuit composed of the MCU controller, LD3320 speech recognition chip and LQ-SYN6288TTS speech synthesis chip in the above embodiment is an independent embedded system, independent of the control board that relies on the Android system, and provides real-time feedback of the sensitive word monitoring results to the user, does not occupy normal speech recognition and playback resources, and can effectively improve the sensitivity, security and authority of the sensitive word monitoring circuit.

[0123] In the above embodiments, although the steps are numbered, such as S1, S2, etc., these are only specific embodiments given in the present application. Those skilled in the art may adjust the execution order of S1, S2, etc. according to actual conditions, which is also within the scope of protection of the present invention. It can be understood that in some embodiments, some or all of the above embodiments may be included.

[0124] like Figure 6 As shown, an apparatus 10 for filtering sensitive words in speech recognition input provided by an embodiment of the present invention includes a sensitive word monitoring module 20 and a general speech recognition module 30;

[0125] The sensitive word monitoring module 20 is used to call the sensitive word monitoring circuit to perform sensitive word monitoring processing on the input voice information to obtain the sensitive word monitoring result corresponding to the voice information;

[0126] The general voice recognition module 30 is used to perform voice recognition processing on the voice information to obtain a voice recognition result corresponding to the voice information; and perform sensitive word filtering on the voice recognition result according to the sensitive word monitoring result to obtain a voice recognition result after sensitive word filtering.

[0127] The device adopting the above structure, while performing voice recognition processing on the input voice information, adopts an independent sensitive word monitoring circuit to realize the sensitive word monitoring function, so as to obtain sensitive words from the input voice information in real time, which can effectively reduce the consumption of software and hardware resources of the main function of voice recognition by sensitive word filtering, is conducive to the normal realization of the main function of voice recognition, and reduces the influence of network delay on the sensitive word screening effect, thereby enhancing the reliability and stability of sensitive word filtering.

[0128] Optionally, in one embodiment, the sensitive word monitoring module 20 includes a first control module; the first control module is used to call the voice recognition chip in the sensitive word monitoring circuit to perform sensitive word monitoring processing on the input voice information to obtain a sensitive word monitoring result corresponding to the voice information.

[0129] Preferably, the first control module is also used to call the voice monitoring microphone to obtain first voice information from the voice input source, based on a preset sensitive word list, call the voice recognition chip to obtain sensitive word information from the first voice information, and generate the sensitive word monitoring result according to the sensitive word information.

[0130] Preferably, the general speech recognition module 30 includes a second control module; the second control module is used to call the speech recognition main microphone to obtain second speech information from the speech input source, and convert the second speech information into corresponding text data based on the speech recognition software development kit to obtain the speech recognition result.

[0131] Optionally, in one embodiment, the sensitive word monitoring result includes a sensitive word index, and the general speech recognition module 30 also includes a processing module; the processing module is used to delete the sensitive words corresponding to the sensitive word index from the speech recognition result to obtain the speech recognition result after the sensitive words are filtered.

[0132] Preferably, the first control module is also used to call the speech synthesis chip in the sensitive word monitoring circuit to generate sensitive word alarm voice information containing the sensitive word information according to the sensitive word information in the sensitive word monitoring result, and to issue a sensitive word alarm voice signal based on the sensitive word alarm voice information.

[0133] Preferably, the first control module is further used to send the sensitive word monitoring result to an application host.

[0134] Those skilled in the art know that the present invention can be implemented as an apparatus, a method or a computer program product. Therefore, the present disclosure can be specifically implemented in the following forms, namely: it can be complete hardware, it can be complete software (including firmware, resident software, microcode, etc.), or it can be a combination of hardware and software, which is generally referred to as a "circuit", "module" or "system" herein. In addition, in some embodiments, the present invention can also be implemented in the form of a computer program product in one or more computer-readable media, and the computer-readable medium contains computer-readable program code.

[0135] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0136] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A method for filtering sensitive words in speech recognition input, characterized in that: include: Performing voice recognition processing on the input voice information to obtain a voice recognition result corresponding to the voice information, and simultaneously performing sensitive word monitoring processing on the input voice information using a sensitive word monitoring circuit to obtain a sensitive word monitoring result corresponding to the voice information; According to the sensitive word monitoring result, the speech recognition result is filtered for sensitive words to obtain a speech recognition result after sensitive word filtering; The sensitive word monitoring result includes a sensitive word index, and the sensitive word filtering is performed on the speech recognition result according to the sensitive word monitoring result to obtain the speech recognition result after the sensitive word filtering, including: deleting the sensitive word corresponding to the sensitive word index from the speech recognition result to obtain the speech recognition result after the sensitive word filtering; The method comprises deleting the sensitive word corresponding to the sensitive word index from the speech recognition result to obtain the speech recognition result after the sensitive word filtering, including: for each sensitive word screened from the input speech information, using a sensitive word monitoring circuit to look up a table to obtain a corresponding serial number, summarizing all the serial numbers obtained by looking up the table to obtain a sensitive word index, sending the sensitive word index from the sensitive word monitoring circuit to an execution module for deleting sensitive words from the speech recognition result, and the execution module looking up the table based on the sensitive word index to obtain the corresponding sensitive word for sensitive word filtering.

2. The method according to claim 1, characterized in that The process of obtaining the sensitive word monitoring result includes: The voice recognition chip in the sensitive word monitoring circuit is used to perform sensitive word monitoring processing on the input voice information to obtain a sensitive word monitoring result corresponding to the voice information.

3. The method according to claim 2, characterized in that The method of using the voice recognition chip in the sensitive word monitoring circuit to perform sensitive word monitoring processing on the input voice information to obtain a sensitive word monitoring result corresponding to the voice information includes: A voice monitoring microphone is used to obtain first voice information from a voice input source, and based on a preset sensitive word list, the voice recognition chip is used to obtain sensitive word information from the first voice information, and the sensitive word monitoring result is generated according to the sensitive word information.

4. The method according to claim 3, characterized in that The performing speech recognition processing on the input speech information to obtain a speech recognition result corresponding to the speech information includes: The second voice information is acquired from the voice input source using a voice recognition main microphone, and the second voice information is converted into corresponding text data based on a voice recognition software development kit to obtain the voice recognition result.

5. The method according to any one of claims 1 to 4, characterized in that The step of performing sensitive word monitoring processing on the input voice information by using the sensitive word monitoring circuit to obtain a sensitive word monitoring result corresponding to the voice information further includes: According to the sensitive word information in the sensitive word monitoring result, a speech synthesis chip in the sensitive word monitoring circuit is used to generate sensitive word alarm voice information containing the sensitive word information, and a sensitive word alarm voice signal is issued based on the sensitive word alarm voice information.

6. The method according to any one of claims 1 to 4, characterized in that Before filtering the speech recognition result for sensitive words according to the sensitive word monitoring result to obtain the speech recognition result after the sensitive word filtering, the method further includes: The sensitive word monitoring result is sent to the application host.

7. A device for filtering sensitive words in speech recognition input, characterized in that: Including sensitive word monitoring module and general speech recognition module; The sensitive word monitoring module is used to call the sensitive word monitoring circuit to perform sensitive word monitoring processing on the input voice information to obtain the sensitive word monitoring result corresponding to the voice information; The general speech recognition module is used to perform speech recognition processing on the speech information to obtain a speech recognition result corresponding to the speech information; According to the sensitive word monitoring result, the speech recognition result is filtered for sensitive words to obtain a speech recognition result after sensitive word filtering; The sensitive word monitoring result includes a sensitive word index, and the general speech recognition module is specifically used to delete the sensitive word corresponding to the sensitive word index from the speech recognition result to obtain the speech recognition result after the sensitive word is filtered; The general speech recognition module is specifically used to use a sensitive word monitoring circuit to look up a table for each sensitive word screened from the input speech information to obtain a corresponding serial number, summarize all the serial numbers obtained from the table lookup to obtain a sensitive word index, and send the sensitive word index from the sensitive word monitoring circuit to an execution module for deleting sensitive words from the speech recognition results. The execution module looks up the table based on the sensitive word index to obtain the corresponding sensitive word for sensitive word filtering.

8. A control panel for filtering sensitive words in voice recognition input, characterized in that: The control panel is used to perform voice recognition processing on the input voice information to obtain a voice recognition result corresponding to the voice information; The control board is further used to receive a sensitive word monitoring result after the voice information is monitored for sensitive words from the controller, and perform sensitive word filtering on the voice recognition result according to the sensitive word monitoring result to obtain a voice recognition result after sensitive word filtering; The sensitive word monitoring result includes a sensitive word index, and the control panel is specifically used to delete the sensitive word corresponding to the sensitive word index from the speech recognition result to obtain the speech recognition result after the sensitive word is filtered; The control board is specifically used to use a sensitive word monitoring circuit to look up a table to obtain a corresponding serial number for each sensitive word screened from the input voice information, summarize all the serial numbers obtained from the table lookup to obtain a sensitive word index, and send the sensitive word index from the sensitive word monitoring circuit to an execution module for deleting sensitive words from the voice recognition results. The execution module looks up the table based on the sensitive word index to obtain the corresponding sensitive word for sensitive word filtering.

9. A device for filtering sensitive words in speech recognition input, characterized in that: The control panel according to claim 8 further comprises a controller; The controller is used to perform sensitive word monitoring processing on the input voice information using the sensitive word monitoring circuit, obtain the sensitive word monitoring result corresponding to the voice information, and send the sensitive word monitoring result to the control board.

Images