A communication lead analysis reminder method and system

Abstract

The application relates to a communication clue analysis reminding method and system, and relates to the technical field of communication clue analysis, which comprises the following steps: collecting audio recording information; pre-processing the audio recording information to extract speech recording information and environment recording information; extracting a person's voiceprint feature and recording text information according to the speech recording information; inputting the person's voiceprint feature into a preset voice recognition model to identify and obtain person identity information; combining the person identity information, the recording text information and the environment recording information to generate a communication clue probability value; when the communication clue probability value is greater than a preset probability reference value, generating communication warning information according to the person identity information and the recording text information, and outputting the communication warning information. The application has the effect of improving the warning accuracy of communication clues in the incoming call communication scene.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of communication clue analysis, in particular to a communication clue analysis reminding method and system. BACKGROUND

[0002] Communication clue analysis refers to a process of systematically collecting, sorting, correlating and judging various data and information generated in communication to discover potential rules, abnormal behaviors or key clues.

[0003] Currently, in the communication scenarios of report hotline, government service hotline, financial customer service, etc., when communication clue analysis is performed, audio recording, time, location and other information are generally collected in the incoming call, and after preprocessing such as noise reduction, voiceprint features are extracted and compared with the database to quickly identify the identity of the caller; at the same time, the call content is converted into text for analysis, and once the system judges that the clue is repeated or suspicious information, it will be automatically marked and sent to remind, in order to achieve the purpose of early warning and improving the disposal efficiency.

[0004] When voiceprint recognition is used in the incoming call communication scenario, the voiceprint recognition is easily affected by environmental noise, resulting in a decrease in recognition accuracy and possible misjudgment, thereby making it inconvenient to accurately warn the communication clue. SUMMARY

[0005] In order to improve the warning accuracy of communication clues in the incoming call communication scenario, the present application provides a communication clue analysis reminding method and system.

[0006] In a first aspect, the present application provides a communication clue analysis reminding method, which adopts the following technical solution:

[0007] A communication clue analysis reminding method, comprising:

[0008] S1: collecting audio recording information;

[0009] S2: preprocessing according to the audio recording information to extract speech recording information and environmental recording information;

[0010] S3: extracting character voiceprint features and recording text information according to the speech recording information;

[0011] S4: inputting the character voiceprint features into a preset voice recognition model to identify and obtain character identity information;

[0012] S5: combining the character identity information, the recording text information and the environmental recording information to generate a communication clue probability value;

[0013] S6: When the communication clue probability value is greater than a preset probability reference value, generating a communication early warning information according to the person identity information and the audio text information, and outputting the communication early warning information.

[0014] By adopting the technical solution, the voiceprint, the text, and the environmental information are fused, the communication clue probability value is generated through three-dimensional analysis, the threshold value is exceeded to accurately warn, the risk subject and the content are quickly located, the redundant warning is avoided, the overallness of the clue analysis and the response efficiency are improved, the influence of the environmental noise is reduced, and the warning accuracy of the communication clue in the incoming call communication scene is improved.

[0015] Optionally, the method for generating the communication clue probability value comprises:

[0016] S51: determining historical text information and selected text information according to the person identity information and the audio text information;

[0017] S52: comparing the selected text information with the historical text information to determine a text coincidence degree;

[0018] S53: generating a reference coincidence degree in combination with the environmental audio information and the person identity information;

[0019] S54: determining whether the text coincidence degree is greater than the reference coincidence degree;

[0020] S55: if yes, calculating a difference value of the text coincidence degree and the reference coincidence degree as a coincidence deviation value;

[0021] S56: generating an identity estimation probability value in combination with the coincidence deviation value and the person identity information, and taking the identity estimation probability value as the communication clue probability value;

[0022] S57: if no, outputting a preset normal probability value as the communication clue probability value.

[0023] By adopting the technical solution, the text coincidence degree and the reference coincidence degree are determined, and whether the text coincidence degree is greater than the reference coincidence degree is judged. When the text coincidence degree is greater than the reference coincidence degree, the identity estimation probability value is generated in combination with the coincidence deviation value and the person identity information, and taken as the communication clue probability value. When the text coincidence degree is not greater than the reference coincidence degree, the preset normal probability value is outputted and taken as the communication clue probability value, so that the accuracy of the obtained communication clue probability value is improved.

[0024] Optionally, the method for determining the historical text information and the selected text information comprises:

[0025] S511: based on the person identity information, calling an identity number value and historical single-identity text information;

[0026] S512: determining whether the identity number value is only one;

[0027] S513: if yes, directly taking the historical single-identity text information as the historical text information and taking the recording text information as the selected text information;

[0028] S514: if no, calling simultaneous text information and time-sharing text information based on the recording text information;

[0029] S515: combining the simultaneous text information and the time-sharing text information to determine a simultaneous text proportion value;

[0030] S516: selecting the character identity information according to the simultaneous text proportion value to obtain selected identity information, taking the historical single-identity text information corresponding to the selected identity information as the historical text information, and taking the recording text information corresponding to the selected identity information as the selected text information.

[0031] By adopting the above technical solution, the text information is determined according to the number of identities and the scene, the simultaneous text information and the time-sharing text information are called when there are multiple identities, the simultaneous text proportion value is determined, and the character identity information is selected according to the simultaneous text proportion value, so as to ensure that the text is strongly related to the target identity, thereby improving the accuracy of the obtained historical text information and the selected text information, and providing high-quality data for subsequent analysis.

[0032] Optionally, the selection method of the selected identity information includes:

[0033] S5161: when the simultaneous text proportion values are all greater than a preset reference proportion value, calling a simultaneous text intensity value from the speech recording information according to the simultaneous text information;

[0034] S5162: calling a single-identity intensity interval based on the character identity information;

[0035] S5163: combining the simultaneous text intensity value and the single-identity intensity interval to determine a single-identity intensity deviation value;

[0036] S5164: combining the single-identity intensity deviation value and the simultaneous text intensity value to determine a simultaneous text intensity reference value;

[0037] S5165: selecting a larger value according to the simultaneous text intensity reference value, and taking the corresponding character identity information as the selected identity information.

[0038] By adopting the technical scheme, when multiple identities and simultaneous text proportion reach the standard, the simultaneous text intensity reference value is determined by using the simultaneous text intensity value and the single-identity intensity interval, and then the identity information corresponding to the larger value is selected as the selected identity information according to the simultaneous text intensity reference value, so as to avoid subjective selection and improve the objectivity of identity selection.

[0039] Optionally, the selection method of the selected identity information further includes:

[0040] S5166: When the simultaneous text proportion value is greater than the preset reference proportion value, the meaning contact value is determined according to the split-time text information.

[0041] S5167: Whether the meaning contact value is greater than the preset contact reference value is determined.

[0042] S5168: If yes, all the character identity information is selected as the selected identity information.

[0043] S5169: If no, the character identity information is selected according to the simultaneous text proportion value and is used as the selected identity information.

[0044] By adopting the technical scheme, when multiple identities and simultaneous text proportion do not reach the standard, the identity is selected according to the meaning contact value, comprehensive and accurate selection is considered, and potential related subjects are avoided to be missed or misselected.

[0045] Optionally, the generation method of the reference coincidence degree includes:

[0046] S531: The identity action range is called based on the character identity information.

[0047] S532: The range site type is determined according to the identity action range.

[0048] S533: The environment recording type and the environment recording intensity value are determined according to the environment recording information.

[0049] S534: Whether the environment recording type belongs to the preset environment special type is determined.

[0050] S535: If yes, the environment estimation type is determined according to the environment recording type.

[0051] S536: The environment estimation coincidence degree is generated by combining the environment estimation type, the range site type and the environment recording intensity value, and the environment estimation coincidence degree is used as the reference coincidence degree.

[0052] S537: If no, the site estimation coincidence degree is determined according to the range site type, and the site estimation coincidence degree is used as the reference coincidence degree.

[0053] By adopting the technical scheme, the reference coincidence degree is generated according to the judgment result of whether the environment recording type belongs to the preset special environment type, and the accuracy of the obtained reference coincidence degree is improved.

[0054] Optionally, the method for determining the environment recording type and the environment recording intensity value comprises:

[0055] S5331: retrieving an environment recording time point, a single-time recording frequency value and a single-time recording intensity value based on the environment recording information;

[0056] S5332: performing curve analysis on the single-time recording frequency value and the environment recording time point to form a recording frequency change curve;

[0057] S5333: retrieving a frequency change interval time value and a change center frequency value based on the recording frequency change curve;

[0058] S5334: performing curve analysis on the single-time recording intensity value and the environment recording time point to form a recording intensity change curve;

[0059] S5335: determining a frequency estimation type in combination with the frequency change interval time value, the change center frequency value, the environment recording time point and the recording intensity change curve, and taking the frequency estimation type as the environment recording type;

[0060] S5336: determining a time intensity influence value according to the environment recording time point;

[0061] S5337: determining a recording intensity selected value in combination with the recording intensity change curve and the time intensity influence value, and taking the recording intensity selected value as the environment recording intensity value.

[0062] By adopting the technical scheme, the environment recording time point, the single-time recording frequency value and the single-time recording intensity value are subjected to curve analysis to generate the recording frequency change curve and the recording intensity change curve, and the frequency change interval time value, the change center frequency value and the time intensity influence value are combined to finally accurately determine the frequency estimation type and the recording intensity selected value, thereby providing high-precision environment recording type and environment recording intensity value basis for generating reliable reference coincidence degree, and significantly improving the accuracy of subsequent communication clue analysis.

[0063] Optionally, the method for generating the frequency estimation type comprises:

[0064] S53351: determining an interval frequency similarity in combination with the frequency change interval time value and the change center frequency value;

[0065] S53352: retrieving an intensity variation interval time value and a variation center intensity value based on the sound recording intensity variation curve;

[0066] S53353: determining an interval intensity decay value by combining the intensity variation interval time value and the variation center intensity value;

[0067] S53354: determining an echo type estimation value by combining the environment sound recording time point, the interval frequency similarity and the interval intensity decay value;

[0068] S53355: determining whether the echo type estimation value is greater than a preset echo type reference value;

[0069] S53356: if yes, outputting a preset echo estimation type as the frequency estimation type;

[0070] S53357: if no, matching and determining a scene estimation type by combining the variation center frequency value and the variation center intensity value, and taking the scene estimation type as the frequency estimation type.

[0071] By using the above technical solution, the interval frequency similarity and the interval intensity decay value are determined to determine the echo type estimation value, so as to preferentially identify and strip the echo type, thereby ensuring the judgment accuracy of the real scene estimation type in a complex acoustic environment, greatly improving the reliability and anti-interference ability of the environment sound recording type identification result, and providing a more pure and reliable data basis for generating an accurate reference coincidence degree.

[0072] Optionally, the environment estimation coincidence degree generation method comprises:

[0073] S5361: selecting a range site type according to the environment estimation type to obtain a selected site type;

[0074] S5362: determining a type estimation position and a type reference intensity value according to the selected site type;

[0075] S5363: determining an environment intensity deviation value by combining the type reference intensity value and the environment sound recording intensity value;

[0076] S5364: determining a distance estimation value according to the environment intensity deviation value;

[0077] S5365: determining an estimation range by combining the type estimation position and the distance estimation value;

[0078] S5366: determining a range estimation coincidence degree by combining the identity action range and the estimation range, and taking the range estimation coincidence degree as the environment estimation coincidence degree.

[0079] By adopting the technical scheme, the selected site type is obtained, the type estimation position and the type reference intensity value are determined, the environmental intensity deviation value is determined, the distance estimation value is determined according to the environmental intensity deviation value, the estimation range is determined, the range estimation coincidence degree is determined according to the identity action range and the estimation range, and the range estimation coincidence degree is taken as the environmental estimation coincidence degree, so that the accuracy of the obtained environmental estimation coincidence degree is improved.

[0080] In a second aspect, the present application provides a communication clue analysis reminding system, which adopts the following technical scheme:

[0081] A communication clue analysis reminding system comprises:

[0082] The acquisition module is configured to acquire audio recording information.

[0083] The memory stores a program for implementing the communication clue analysis reminding method according to any one of the first aspect.

[0084] The processor loads and executes the program stored in the memory.

[0085] In summary, the present application has at least one of the following beneficial technical effects:

[0086] 1. The audio recording information is acquired and extracted to obtain speech recording information and environmental recording information, the voiceprint features and recording text information of the speaker are extracted from the speech recording information, the identity information of the speaker is recognized and obtained, the communication clue probability value is generated according to the identity information of the speaker, the recording text information and the environmental recording information, when the communication clue probability value is greater than a preset probability reference value, the communication warning information is generated according to the identity information of the speaker and the recording text information, and the communication warning information is outputted, so as to reduce the interference of environmental sound and improve the warning accuracy of the communication clue in the incoming call communication scene.

[0087] 2. The historical text information and the selected text information are determined according to the identity information of the speaker and the recording text information, and the text coincidence degree is obtained by comparison, the reference coincidence degree is generated according to the environmental recording information and the identity information of the speaker, and according to the judgment result whether the text coincidence degree is greater than the reference coincidence degree, the identity estimation probability value is analyzed and generated or the normal probability value is outputted as the communication clue probability value, so as to improve the accuracy of the obtained communication clue probability value.

[0088] 3. The method comprises the following steps: calling the identity number value and historical single-identity text information through the character identity information; when the identity number value is only one, directly taking the historical single-identity text information as the historical text information and taking the recording text information as the selected text information; when the identity number value is not only one, calling the simultaneous text information and the split-time text information through the recording text information, determining the simultaneous text proportion value, selecting the character identity information through the simultaneous text proportion value to obtain the selected identity information, taking the historical single-identity text information corresponding to the selected identity information as the historical text information, and taking the recording text information corresponding to the selected identity information as the selected text information, so as to improve the accuracy of the obtained historical text information and the selected text information. BRIEF DESCRIPTION OF DRAWINGS

[0089] Figure 1 It is a method flowchart of communication clue analysis reminding. DETAILED DESCRIPTION

[0090] The application will be described in further detail below with reference to the drawings and embodiments.

[0091] A communication clue analysis reminding method, which comprises the following steps: collecting audio recording information and extracting speech recording information and environmental recording information, so as to analyze the text coincidence degree, the environmental benchmark coincidence degree and the occupation correlation degree for multi-dimensional verification, thereby improving the accuracy of the obtained communication clue probability value.

[0092] REFERENCE Figure 1 The embodiment of the application discloses a communication clue analysis reminding method, which comprises the following steps:

[0093] S1: collecting audio recording information.

[0094] The audio recording information refers to original audio data recorded by an audio collection device in a communication scene.

[0095] The communication scene includes calling, voice interaction, clue feedback and the like. The audio collection device refers to a device for collecting sound, which can be a mobile phone microphone, a fixed telephone recording module, a recording pen, a voice collection terminal and the like. The audio recording information includes human voice such as speech content and environmental sound such as background noise and scene sound effects.

[0096] When in a case reporting hotline, a government service hotline, a financial customer service and the like, the audio recording information is collected by the audio collection device, thereby providing a basic data source for subsequent voiceprint recognition, text extraction and clue analysis.

[0097] S2: pre-processing according to the audio recording information to extract speech recording information and environmental recording information.

[0098] The speech recording information refers to an audio segment in the audio recording information that only contains human voice of the caller, and the environmental recording information refers to an audio segment that reflects a scene background and does not contain human voice.

[0099] Since human voice is usually concentrated in a medium frequency range (300 Hz to 3400 Hz), and environmental sound is generally distributed in other frequency bands. After preprocessing such as cleaning the audio recording information, the audio in the time period in the medium frequency range is intercepted as the speech recording information through a band-pass filter, and the audio in the time period that does not contain human voice is taken as the environmental recording information, facilitating further analysis.

[0100] S3: Extracting a voiceprint feature of a person and recording text information according to the speech recording information.

[0101] The voiceprint feature of the person refers to physiological and behavioral characteristics such as pronunciation habits, formant frequencies, and pitch periods that reflect the uniqueness of an individual, and the recording text information refers to information corresponding to the speech content in the speech recording information converted into a text form.

[0102] The voiceprint feature of the person is obtained by converting the speech recording information into a waterfall spectrogram to extract image features. The voice content in the speech recording is converted into text data through speech-to-text technology, and the structured recording text information is obtained through natural language processing for word segmentation and cleaning, facilitating subsequent use. The voiceprint feature extraction and speech-to-text technology are both existing technologies, and will not be described in detail.

[0103] S4: Inputting the voiceprint feature of the person into a preset voice recognition model to identify and obtain person identity information.

[0104] The voice recognition model refers to an algorithm model that is trained in advance and is used to match identity through voiceprint features. The voice recognition model is constructed based on neural network technology and can store voiceprint feature templates of known persons. The construction of the voice recognition model is an existing technology, and will not be described in detail.

[0105] The extracted voiceprint feature of the person is input into the preset voice recognition model. The voice recognition model compares the voiceprint feature with the built-in voiceprint feature template library to calculate the feature similarity. When the similarity exceeds a preset threshold, the matching is successful and the corresponding person identity information is output. If no similar template is matched, the identity is marked as unknown, and the identity recognition process is completed. The voiceprint feature template library pre-stores voiceprint features of known persons.

[0106] S5: Generating a communication clue probability value by combining the person identity information, the recording text information, and the environmental recording information.

[0107] The communication clue probability value is a probability value corresponding to the effectiveness of the current communication clue, the correlation with historical clues, and the existence of abnormalities.

[0108] The communication clue probability value is generated by combining and analyzing the character identity information, the audio text information, and the environmental audio information, thereby facilitating subsequent judgment of whether to trigger an early warning.

[0109] In order to further ensure the rationality of the communication clue probability value, the communication clue probability value needs to be further analyzed and calculated separately, and the specific steps are as follows.

[0110] The method for generating the communication clue probability value comprises the following steps:

[0111] S51: Determine the historical text information and the selected text information according to the character identity information and the audio text information.

[0112] The historical text information refers to the past stored text data associated with the current character identity information. The selected text information refers to the text content selected from the current audio text information.

[0113] The historical text information and the selected text information are determined by analyzing the character identity information and the audio text information, thereby facilitating subsequent use.

[0114] In order to further ensure the rationality of the historical text information and the selected text information, the historical text information and the selected text information need to be further analyzed and calculated separately, and the specific steps are as follows.

[0115] The method for determining the historical text information and the selected text information comprises the following steps:

[0116] S511: Retrieve the identity number value and the historical single-identity text information based on the character identity information.

[0117] The identity number value refers to the number corresponding to the character identity information. The historical single-identity text information refers to the historical text data corresponding to a single identity.

[0118] The identity number value is obtained by counting the character identity information, and the historical single-identity text information is obtained by querying a preset identity history database based on the character identity information, thereby facilitating subsequent use.

[0119] The identity history database pre-stores a comparison table of different character identities and corresponding historical texts, and the identity history database is obtained by storing text data of different character identities in real time.

[0120] S512: Determine whether the identity number value is only one. If yes, execute S513; if no, execute S514.

[0121] Wherein, by judging whether the identity number value is only one, it is judged whether further selection is needed.

[0122] S513: Directly take the historical single-identity text information as the historical text information, and take the recording text information as the selected text information.

[0123] Wherein, when the identity number value is only one, it means that further selection is not needed at this time, so the historical single-identity text information is directly taken as the historical text information, and the recording text information is taken as the selected text information, thereby facilitating subsequent use.

[0124] S514: Based on the recording text information, call the simultaneous text information and the split-time text information.

[0125] Wherein, the simultaneous text information refers to the text data corresponding to the text when different character identities exist in the same period. The split-time text information refers to the text data unique to different character identities in a period.

[0126] When the identity number value is only one, it means that further selection is needed at this time, so the recording text information corresponding to different character identities is compared according to the time period. When text data exists in the same period, it is extracted and taken as the simultaneous text information, and the text data corresponding to the remaining period is taken as the split-time text information, thereby facilitating further selection.

[0127] S515: Combine the simultaneous text information and the split-time text information to determine the simultaneous text proportion value.

[0128] Wherein, the simultaneous text proportion value refers to the proportion between the text data of the simultaneous text information and the text data of the split-time text information.

[0129] By first counting the time period corresponding to the simultaneous text information and the time period of the split-time text information to obtain the simultaneous text time period and the split-time text time period, the simultaneous text duration value corresponding to the simultaneous text time period and the split-time text duration value corresponding to the split-time text time period are calculated, and then the proportion value between the simultaneous text duration value and the split-time text duration value is calculated to obtain the simultaneous text proportion value, thereby facilitating subsequent selection.

[0130] Since the split-time text information corresponding to different character identities is different, the simultaneous text proportion value corresponding to different character identities is also different.

[0131] S516: Select the identity information of the person according to the simultaneous text proportion value, obtain the selected identity information, and take the historical single-identity text information corresponding to the selected identity information as the historical text information, and take the recording text information corresponding to the selected identity information as the selected text information.

[0132] The selected identity information refers to the identity information corresponding to the selected identity information of the person.

[0133] The selected identity information is obtained by selecting the identity information of the person according to the simultaneous text proportion value, and the historical single-identity text information corresponding to the selected identity information is taken as the historical text information, and the recording text information corresponding to the selected identity information is taken as the selected text information, thereby improving the accuracy of the obtained historical text information and the selected text information.

[0134] In order to further ensure the rationality of the selected identity information, it is necessary to further analyze and calculate the selected identity information separately, which will be described in detail as follows.

[0135] The selection method of the selected identity information includes the following steps:

[0136] S5161: When the simultaneous text proportion value is greater than the preset reference proportion value, the simultaneous text intensity value is retrieved from the speech recording information according to the simultaneous text information.

[0137] The reference proportion value refers to the minimum proportion value corresponding to the simultaneous speech of multiple persons, which is set by the operator in advance. The operator can adjust it according to the scene requirements such as public security and government service.

[0138] The simultaneous text intensity value refers to the quantized value of the voice signal intensity of the voice segment corresponding to the simultaneous text information in the speech recording information, which is used to reflect the clarity and energy of the voice segment.

[0139] When the simultaneous text proportion value is greater than the preset reference proportion value, it means that multiple persons are speaking simultaneously. Therefore, the simultaneous text intensity value corresponding to the simultaneous text information is retrieved from the speech recording information for subsequent use.

[0140] S5162: Retrieve the single-identity intensity interval based on the identity information of the person.

[0141] The single-identity intensity interval refers to the numerical range in which the voice signal intensity of each independent person identity usually falls.

[0142] The single-identity intensity interval is obtained by inputting the identity information of the person into the preset identity history database, which is convenient for subsequent use.

[0143] The identity history database pre-stores a correspondence table of different person identities and corresponding single identity strength intervals, and the identity history database obtains the single identity strength interval of different person identities in real time.

[0144] S5163: Determine a single identity strength deviation value by combining the simultaneous text strength value and the single identity strength interval.

[0145] The single identity strength deviation value refers to a deviation value corresponding to the deviation of the speech signal strength corresponding to a single identity.

[0146] By analyzing the simultaneous text strength value and the single identity strength interval, when the simultaneous text strength value falls within the single identity strength interval, 0 is directly output as the single identity strength deviation value. When the simultaneous text strength value does not fall within the single identity strength interval, the difference between the simultaneous text strength value and the nearest boundary of the single identity strength interval is calculated as the deviation value, and the single identity strength deviation value is finally obtained for subsequent use.

[0147] S5164: Determine a simultaneous text strength reference value by combining the single identity strength deviation value and the simultaneous text strength value.

[0148] The simultaneous text strength reference value refers to a corrected value calculated by combining the current simultaneous text strength value and the single identity strength deviation value.

[0149] By inputting the single identity strength deviation value into the preset strength deviation database to match the strength deviation coefficient value, and calculating the product value between the strength deviation coefficient value and the simultaneous text strength value as the simultaneous text strength reference value, subsequent use is facilitated.

[0150] The strength deviation database pre-stores different single identity strength deviation intervals and corresponding strength deviation coefficient values. The larger the single identity strength deviation value, the smaller the strength deviation coefficient value, and the strength deviation database is pre-set by an operator according to actual needs.

[0151] For example, the strength deviation coefficient value corresponding to the single identity strength deviation interval of 0 to 1 can be set to 1, the strength deviation coefficient value corresponding to the single identity strength deviation interval of 1 to 5 can be set to 0.8, and the strength deviation coefficient value corresponding to the single identity strength deviation interval greater than 5 can be set to 0.6.

[0152] S5165: Select a larger value according to the simultaneous text strength reference value, and take the corresponding person identity information as the selected identity information.

[0153] The method comprises the following steps: comparing the reference values of the simultaneous text intensity, selecting the larger value in the comparison result, and taking the character identity corresponding to the selected simultaneous text intensity as the selected identity information, thereby improving the accuracy of the selected identity information.

[0154] In order to further ensure the rationality of the selected identity information, it is necessary to further analyze and calculate the selected identity information. The specific steps are as follows.

[0155] The selected identity information selection method further comprises the following steps:

[0156] S5166: When the simultaneous text proportion value is not uniform and is greater than the preset reference proportion value, the meaning contact value is determined according to the time-sharing text information.

[0157] The meaning contact value refers to a numerical value quantifying the close degree of the semantic correlation of each time-sharing text information.

[0158] When the simultaneous text proportion value is not uniform and is greater than the preset reference proportion value, it means that multiple people are not speaking at the same time. Therefore, by extracting the core semantic elements such as keywords, event subjects, and appeal types in the time-sharing text information, and using natural language processing techniques such as semantic similarity algorithm and word vector cosine similarity, the matching degree of the core semantic elements is compared, and the score is quantified according to the comparison result. The higher the similarity, the higher the score, and the full score is set to 1 or 100%. The score is the meaning contact value, which is convenient for subsequent use.

[0159] S5167: Determine whether the meaning contact value is greater than the preset contact reference value. If yes, execute S5168; if no, execute S5169.

[0160] The contact reference value refers to the minimum threshold corresponding to the existence of contact. The contact reference value is obtained by pre-inputting by the operator.

[0161] By determining whether the meaning contact value is greater than the preset contact reference value, the selection range of the selected identity information is determined.

[0162] S5168: Take all the character identity information as the selected identity information.

[0163] When the meaning contact value is greater than the preset contact reference value, it means that there is a greater correlation between the speeches of multiple people, so all the character identity information is taken as the selected identity information, thereby improving the accuracy of the selected identity information.

[0164] S5169: Select the character identity information according to the simultaneous text proportion value and take it as the selected identity information.

[0165] When the meaning correlation value is not greater than the preset correlation reference value, it indicates that there is no greater correlation between the speeches of multiple people at this time, so by comparing the simultaneous text proportion values and selecting the maximum value of the simultaneous text proportion values as the selected identity information, the accuracy of the selected identity information obtained is improved.

[0166] S52: Compare the selected text information with the historical text information to determine the text coincidence degree.

[0167] The text coincidence degree refers to the matching degree quantization value of the selected text and the historical text in content.

[0168] By unified cleaning of the selected text information and the historical text information, and performing word segmentation and keyword extraction, and then using cosine similarity, Jaccard similarity coefficient and other text similarity algorithms to compare the coincidence number of the keywords and the semantic matching degree of the core sentences of the two, and then outputting a quantization value of 0 to 1 or 0% to 100% according to the matching degree, the value is the text coincidence degree, thereby facilitating subsequent comparison.

[0169] S53: Combine the environmental recording information and the character identity information to generate a reference coincidence degree.

[0170] The reference coincidence degree refers to the minimum coincidence degree of the selected text and the historical text in content according to the character identity and the environment.

[0171] By combining and analyzing the environmental recording information and the character identity information, the reference coincidence degree is generated, which is convenient for subsequent use.

[0172] In order to further ensure the rationality of the reference coincidence degree, it is necessary to make a further separate analysis and calculation of the reference coincidence degree, which is specifically described as follows.

[0173] The method for generating the reference coincidence degree further includes the following steps:

[0174] S531: Retrieve the identity action range based on the character identity information.

[0175] The identity action range refers to the range of action of the character identity under normal circumstances.

[0176] The identity action range is matched by querying the preset identity history database based on the character identity information, which is convenient for subsequent use.

[0177] The identity history database pre-stores a correspondence table of different personal identities and corresponding historical texts, stores the location in real time when the identity history database communicates through different personal identities, and determines the identity action range according to multiple historical locations. When there is only one, the range of the preset distance around the location is taken as the identity action range. When there are multiple, the ranges of the preset distances around the multiple locations are determined, and then the multiple ranges are connected to form the identity action range. The preset distance is pre-set by the operator according to the actual situation.

[0178] S532: Determine the range site type according to the identity action range.

[0179] The range site type refers to the type of each site in the identity action range, and the range site type includes office sites, shopping malls, train stations, residential areas, and the like.

[0180] By splitting each site included in the identity action range, and taking the type of the split site as the range site type, subsequent use is facilitated.

[0181] S533: Retrieve the environmental recording type and environmental recording intensity value based on the environmental recording information.

[0182] The environmental recording type refers to a label classified according to the source and scene characteristics of the background sound. The environmental recording intensity value refers to a numerical value quantified from the intensity of the sound signal corresponding to the environmental recording information.

[0183] By matching the sound characteristics corresponding to the environmental recording information with the preset sound type database to obtain the environmental recording type, and retrieving the intensity value corresponding to the environmental recording information to obtain the environmental recording intensity value, subsequent use is facilitated.

[0184] The sound type database pre-stores a correspondence table of different sound characteristics and corresponding environmental recording types, and the sound type database is obtained by the operator pre-inputting various sound characteristics and corresponding environmental recording types.

[0185] In order to further ensure the rationality of the environmental recording type and the environmental recording intensity value, it is necessary to make further separate analysis and calculation of the environmental recording type and the environmental recording intensity value. The specific steps are as follows.

[0186] The method for determining the environmental recording type and the environmental recording intensity value includes the following steps:

[0187] S5331: Retrieve the environmental recording time point, single-time recording frequency value, and single-time recording intensity value based on the environmental recording information.

[0188] Among them, the environmental recording time point refers to each specific moment when the audio signal is sampled during the recording of environmental audio information. The single-time recording frequency value refers to the frequency contained in the environmental sound signal at each moment of recording. The single-time recording intensity value refers to the amplitude or energy intensity of the environmental sound signal at each moment of recording.

[0189] Environmental recording information includes environmental recording time points, single-time recording frequency values, and single-time recording intensity values.

[0190] The environmental recording information can be used to retrieve the environmental recording time point, single-time recording frequency value, and single-time recording intensity value for later use.

[0191] S5332: Combine single-time recording frequency values ​​with environmental recording time points to perform curve analysis to form a recording frequency change curve.

[0192] The recording frequency change curve refers to the curve that describes how the recording frequency value at a single time point changes with the environmental recording time point.

[0193] By using the environmental recording time points as the horizontal axis (time axis) and the single-time recording frequency value as the vertical axis (frequency axis), and visually fitting the corresponding data of the two to form a curve, a recording frequency change curve is generated for convenient subsequent use.

[0194] S5333: Retrieve the frequency change interval time value and the center frequency value of the change based on the recording frequency change curve.

[0195] The frequency change interval value refers to the difference between the time points of two adjacent significant frequency changes (such as a sudden increase, decrease, or change in fluctuation pattern) in the recorded frequency change curve. The center frequency value refers to the center frequency within the relatively stable frequency range in the curve before and after each significant frequency change.

[0196] By identifying the recording frequency change curve based on a preset frequency change threshold, the curve portion exceeding the frequency change threshold is located. Then, the frequency change interval value and the center frequency value are retrieved based on the located curve for convenient subsequent use.

[0197] The frequency change threshold is a threshold used to indicate when there is a significant change. The frequency change threshold is preset by the operator.

[0198] S5334: Combine single-time recording intensity values ​​with environmental recording time points to perform curve analysis to form a recording intensity change curve.

[0199] The recording intensity change curve refers to a curve describing the change of the single-time recording intensity value with the change of the environmental recording time point.

[0200] The recording intensity change curve is formed by taking the environmental recording time point as the horizontal axis (time axis) and the single-time recording intensity value as the vertical axis (frequency axis), and visualizing and fitting the corresponding data, so as to facilitate subsequent use.

[0201] S5335: Determine the frequency estimation type based on the frequency change interval time value, the change center frequency value, the environmental recording time point, and the recording intensity change curve, and take the frequency estimation type as the environmental recording type.

[0202] The frequency estimation type refers to a type corresponding to the environmental recording estimation according to the frequency situation.

[0203] By analyzing the frequency change interval time value, the change center frequency value, the environmental recording time point, and the recording intensity change curve, the frequency estimation type is determined, and the frequency estimation type is taken as the environmental recording type, thereby improving the accuracy of the obtained environmental recording type.

[0204] In order to further ensure the rationality of the frequency estimation type, it is necessary to make a further separate analysis and calculation on the frequency estimation type. The specific steps are as follows.

[0205] The generation method of the frequency estimation type includes the following steps:

[0206] S53351: Determine the interval frequency similarity based on the frequency change interval time value and the change center frequency value.

[0207] The interval frequency similarity refers to a similarity index of adjacent frequency change patterns.

[0208] The frequency time interval similarity is obtained by calculating the relative difference rate between adjacent frequency change interval time values, and then subtracting the frequency time interval relative difference rate from 1. The center frequency similarity is obtained by calculating the relative difference rate between adjacent change center frequency values, and then subtracting the center frequency relative difference rate from 1. Then, the frequency time interval similarity and the center frequency similarity are weighted and averaged (usually the weights are 50% each), and the result is the interval frequency similarity.

[0209] S53352: Retrieve the intensity change interval time value and the change center intensity value based on the recording intensity change curve.

[0210] The intensity change interval time value refers to the difference between the time points at which two adjacent significant intensity changes (such as sudden intensity increase, decrease, or change in fluctuation mode) occur in the recording intensity change curve. The change center intensity value refers to the center intensity in the relatively stable intensity interval before and after each significant intensity change.

[0211] By identifying the recording intensity change curve according to the preset intensity change threshold, the curve part corresponding to the intensity change threshold is located, and then the intensity change interval time value and the change center intensity value are retrieved according to the located curve, which is convenient for subsequent use.

[0212] The intensity change threshold refers to a threshold corresponding to a significant change, which is set by the operator in advance.

[0213] S53353: Determine the interval intensity decay value by combining the intensity change interval time value and the change center intensity value.

[0214] The interval intensity decay value refers to the change rate and trend of the intensity of the environmental sound in the time dimension.

[0215] The intensity change amount is calculated by calculating the difference between adjacent change center intensity values, and the interval intensity decay value is calculated by calculating the quotient of the intensity change amount and the intensity change interval time value.

[0216] S53354: Determine the echo type estimate value by combining the environmental recording time point, the interval frequency similarity, and the interval intensity decay value.

[0217] The echo type estimate value refers to an index of the possibility and degree of conforming to the echo characteristics (value range 0-1, 1 represents a strong echo tendency, and 0 represents no echo tendency).

[0218] The time falling assignment is obtained by analyzing the falling of the environmental recording time point into the preset echo-prone period, and the echo type estimate value is obtained by weighted calculation of the time falling assignment, the interval frequency similarity, and the interval intensity decay value, which is convenient for subsequent use.

[0219] The echo-prone period is set by the operator in advance. For example, the echo-prone period can be set to 21:00-09:00.

[0220] S53355: Determine whether the echo type estimate value is greater than the preset echo type reference value. If yes, perform S53356; if no, perform S53357.

[0221] The echo type reference value refers to a reference value corresponding to the echo feature. The echo type reference value is obtained by pre-input of an operator.

[0222] The echo type reference value refers to a reference value corresponding to the echo feature. The echo type reference value is obtained by pre-input of an operator.

[0223] The echo type reference value refers to a reference value corresponding to the echo feature. The echo type reference value is obtained by pre-input of an operator.

[0224] The echo type reference value refers to a reference value corresponding to the echo feature. The echo type reference value is obtained by pre-input of an operator.

[0225] The echo type reference value refers to a reference value corresponding to the echo feature. The echo type reference value is obtained by pre-input of an operator.

[0226] The echo type reference value refers to a reference value corresponding to the echo feature. The echo type reference value is obtained by pre-input of an operator.

[0227] The echo type reference value refers to a reference value corresponding to the echo feature. The echo type reference value is obtained by pre-input of an operator.

[0228] The echo type reference value refers to a reference value corresponding to the echo feature. The echo type reference value is obtained by pre-input of an operator.

[0229] The echo type reference value refers to a reference value corresponding to the echo feature. The echo type reference value is obtained by pre-input of an operator.

[0230] S5336: Determine the time intensity influence value according to the environment recording time point.

[0231] The echo type reference value refers to a reference value corresponding to the echo feature. The echo type reference value is obtained by pre-input of an operator.

[0232] The echo type reference value refers to a reference value corresponding to the echo feature. The echo type reference value is obtained by pre-input of an operator.

[0233] The echo type reference value refers to a reference value corresponding to the echo feature. The echo type reference value is obtained by pre-input of an operator.

[0234] For example, the time-intensity influence database can set the time-intensity influence value as 1.2 when the environmental recording time point is 00:00-06:00. The time-intensity influence value can be set as 0.7 when the environmental recording time point is 06:00-09:00. The time-intensity influence value can be set as 1.0 when the environmental recording time point is 09:00-17:00. The time-intensity influence value can be set as 0.6 when the environmental recording time point is 17:00-20:00. The time-intensity influence value can be set as 1.1 when the environmental recording time point is 20:00-24:00.

[0235] S5337: Determine the recording intensity selection value by combining the recording intensity variation curve and the time-intensity influence value, and take the recording intensity selection value as the environmental recording intensity value.

[0236] The recording intensity selection value refers to the intensity value corresponding to the intensity selected and adjusted from the recording intensity variation curve.

[0237] By adjusting the single-time recording intensity value corresponding to each environmental recording time point from the recording intensity variation curve, calculating the product value between each single-time recording intensity value and the time-intensity influence value, obtaining the single-time intensity correction value, calculating the average value between the single-time intensity correction values, taking the calculation result as the recording intensity selection value, and taking the recording intensity selection value as the environmental recording intensity value, the accuracy of the obtained environmental recording intensity value is improved.

[0238] S534: Determine whether the environmental recording type belongs to a preset environmental special type. If yes, perform S535; if no, perform S537.

[0239] The environmental special type refers to an environmental recording category with high interference, high sensitivity, or specific business association. The environmental special type includes train rail collision sound, continuous mechanical roar sound, home appliance explanation sound, etc.

[0240] By determining whether the environmental recording type belongs to a preset environmental special type, it is determined whether the reference coincidence degree can be analyzed according to the environmental condition.

[0241] S535: Determine the environmental estimation type according to the environmental recording type.

[0242] The environmental estimation type refers to the type of the environment estimated according to the environmental recording.

[0243] When the environmental recording type belongs to a preset environmental special type, it means that the reference coincidence degree can be analyzed according to the environmental condition at this time, so the environmental recording type is input into the preset environmental type estimation database to match the environmental estimation type, which is convenient for subsequent use.

[0244] Different environment recording types correspond to different environment estimation types, and the environment type estimation database pre-stores a comparison table of different environment recording types and corresponding environment estimation types, and the environment type estimation database is pre-set by an operator for each environment recording type, environment estimation type, location and intensity value.

[0245] For example, when the environment recording type is a train track collision sound, the environment estimation type is the surrounding of the train line. When the environment recording type is a continuous mechanical roar, the environment estimation type is a factory. When the environment recording type is an appliance explanation sound, the environment estimation type is a shopping mall.

[0246] S536: Combine the environment estimation type, the range site type, and the environment recording intensity value to generate an environment estimation coincidence degree, and take the environment estimation coincidence degree as the reference coincidence degree.

[0247] The environment estimation coincidence degree refers to the coincidence degree corresponding to the matching of the current environment and the character identity adaptation site after estimation.

[0248] By combining and analyzing the environment estimation type, the range site type, and the environment recording intensity value, the environment estimation coincidence degree is generated, and the environment estimation coincidence degree is taken as the reference coincidence degree, thereby improving the accuracy of the obtained reference coincidence degree.

[0249] In order to further ensure the rationality of the environment estimation coincidence degree, it is necessary to make a further separate analysis and calculation of the environment estimation coincidence degree. The specific steps are as follows.

[0250] The method for generating the environment estimation coincidence degree further includes the following steps:

[0251] S5361: Select the range site type according to the environment estimation type to obtain a selected site type.

[0252] The selected site type refers to the type corresponding to the selected range site type.

[0253] By selecting the type consistent with the environment estimation type from the range site type as the selected site type, it is convenient for subsequent use.

[0254] S5362: Determine the type estimation position and the type reference intensity value according to the selected site type.

[0255] The type estimation position refers to the position of the selected site type, and the type reference intensity value refers to the sound intensity standard range corresponding to the environmental sound emitted by the selected site type.

[0256] The field type is input into the preset environment type estimation database to match the type estimation position and the type reference intensity value, facilitating subsequent use.

[0257] The environment type estimation database also pre-stores a comparison table of different selected field types and corresponding type estimation positions and type reference intensity values.

[0258] For example, when the environment estimation type is the surrounding of a train line, the type estimation position is a position point of the surrounding of the train line, the sound intensity interval of the surrounding of the train line is 85 dB to 105 dB, and the type reference intensity value is the interval midpoint 95 dB. When the environment estimation type is a factory, the type estimation position is the position of the factory, the sound intensity interval of the position of the factory is 70 dB to 110 dB, and the type reference intensity value is the interval midpoint 90 dB.

[0259] S5363: Determine the environment intensity deviation value by combining the type reference intensity value and the environment recording intensity value.

[0260] The environment intensity deviation value refers to the deviation value corresponding to the deviation of the sound intensity of the environment.

[0261] The difference between the type reference intensity value and the environment recording intensity value is calculated, and the calculation result is used as the environment intensity deviation value, facilitating subsequent use.

[0262] S5364: Determine the distance estimation value according to the environment intensity deviation value.

[0263] The distance estimation value refers to the distance value corresponding to the deviation of the sound intensity.

[0264] The product value between the environment intensity deviation value and the preset deviation distance coefficient is calculated, and the calculation result is used as the distance estimation value, facilitating subsequent use.

[0265] The deviation distance coefficient is a coefficient for converting the environment intensity deviation value into the distance estimation value, and the deviation distance coefficient is pre-set by the operator according to actual needs.

[0266] S5365: Determine the estimation range by combining the type estimation position and the distance estimation value.

[0267] The estimation range refers to the range corresponding to the estimated position of the dialogue character.

[0268] The type estimation position is taken as the center, the distance estimation value is taken as the radius, and the range thus delimited is used as the estimation range, facilitating subsequent use.

[0269] S5366: Combining the identity action range and the estimated range, determining the range estimated coincidence degree, and taking the range estimated coincidence degree as the environment estimated coincidence degree.

[0270] The range estimated coincidence degree refers to the coincidence degree estimated according to the coincidence of the identity action range and the estimated range.

[0271] By analyzing the coincidence between the identity action range and the estimated range, when there is coincidence, 50% is output at this time and taken as the range estimated coincidence degree, and when there is no coincidence, the shortest distance between the identity action range and the estimated range is calculated, the range longest distance corresponding to the identity action range is calculated, the ratio value between the shortest distance and the range longest distance is calculated, the ratio value is input to the preset range estimated coincidence database to match the range estimated coincidence degree, and the range estimated coincidence degree is taken as the environment estimated coincidence degree, facilitating subsequent use.

[0272] The smaller the ratio value, the closer the range estimated coincidence degree to 50%, and the larger the ratio value, the smaller the range estimated coincidence degree. The range coincidence probability database pre-stores a comparison table of different ratio intervals and corresponding range estimated coincidence degrees, and the range coincidence probability database is pre-set by an operator according to demand.

[0273] For example, when the ratio interval is 0 to 0.3, the range estimated coincidence degree is 50%; when the ratio interval is 0.3 to 0.7, the range estimated coincidence degree is 40%; when the ratio interval is 0.7 to 1.5, the range estimated coincidence degree is 30%; and when the ratio interval is greater than 1.5, the range estimated coincidence degree is 10%.

[0274] S537: According to the range site type, determining the site estimated coincidence degree, and taking the site estimated coincidence degree as the reference coincidence degree.

[0275] The site estimated coincidence degree refers to the degree coincidence corresponding to the estimation according to the range site type. Different range site types correspond to different site estimated coincidence degrees.

[0276] When the environment recording type does not belong to the preset environment special type, it means that the reference coincidence degree cannot be analyzed according to the environment at this time, so the range site type is input to the preset site estimation database to match the site estimated coincidence degree, and the site estimated coincidence degree is taken as the reference coincidence degree, thereby improving the accuracy of the obtained reference coincidence degree.

[0277] The site estimation database pre-stores a comparison table of different range site types and corresponding site estimated coincidence degrees, and the site estimation database is pre-set by an operator according to actual demand.

[0278] For example, when the range site type is an office site, the site estimated coincidence degree is 50%; when the range site type is a shopping mall, the site estimated coincidence degree is 40%; and when the range site type is a railway station, the site estimated coincidence degree is 30%.

[0279] S54: Determine whether the text coincidence degree is greater than the reference coincidence degree. If yes, perform S55; if no, perform S57.

[0280] Wherein, by judging whether the text coincidence degree is greater than the reference coincidence degree, it is determined whether there is an anomaly.

[0281] S55: Calculate the difference between the text coincidence degree and the reference coincidence degree as the coincidence deviation value.

[0282] Wherein, the coincidence deviation value refers to the deviation value corresponding to the deviation of the coincidence degree.

[0283] When the text coincidence degree is greater than the reference coincidence degree, it means that there is an anomaly at this time, so the difference between the text coincidence degree and the reference coincidence degree is calculated, and the calculation result is used as the coincidence deviation value for subsequent use.

[0284] S56: Combine the coincidence deviation value and the character identity information to generate an identity estimation probability value as a communication clue probability value.

[0285] Wherein, the identity estimation probability value refers to the probability value of determining the existence of an anomaly according to the coincidence between the character identity and the current communication clue.

[0286] By combining and analyzing the coincidence deviation value and the character identity information, an identity estimation probability value is generated as a communication clue probability value, improving the accuracy of the obtained communication clue probability value.

[0287] In order to further ensure the rationality of the identity estimation probability value, it is necessary to make a further separate analysis and calculation of the identity estimation probability value, which will be described in detail by the following steps.

[0288] The method for generating the identity estimation probability value further includes the following steps:

[0289] S561: Retrieve recording source information based on audio recording information.

[0290] Wherein, the recording source information refers to the trigger scene information corresponding to the recording collection source. The recording source information can be a report hotline, a government service hotline, a financial customer service, etc. The trigger scene corresponding to the audio recording information is retrieved as the recording source information for subsequent use.

[0291] S562: Determine the character reference occupation information according to the recording source information.

[0292] The character reference occupation information refers to occupation information inferred according to a recording source.

[0293] Different recording source information corresponds to different character reference occupation information. When the recording source information is a case reporting hotline, the character reference occupation information is all occupations. When the recording source information is a government service hotline, the corresponding occupation is determined according to the government category. When the recording source information is a financial customer service, the character reference occupation information is a financial related occupation.

[0294] S563: Retrieve identity occupation information based on character identity information.

[0295] The identity occupation information refers to occupation information corresponding to the character identity. The occupation corresponding to the character identity is retrieved as the identity occupation information, which is convenient for subsequent use.

[0296] S564: Determine occupation relevance by combining identity occupation information and character reference occupation information.

[0297] The occupation relevance refers to the correlation degree corresponding to the occupation.

[0298] The identity occupation information and the character reference occupation information are matched and analyzed by accurate matching degree, large category matching degree, and business association degree, and weighted calculation is performed, so as to obtain the occupation relevance.

[0299] For example, when the identity occupation information is an individual industrial and commercial household, and the character reference occupation information is also an individual industrial and commercial household, the occupation relevance is 100%. When the identity occupation information is an APP planning manager, and the character reference occupation information is an APP operation manager, the occupation relevance is 70%.

[0300] S565: Generate occupation estimation probability value by combining occupation relevance and coincidence deviation value, and take the occupation estimation probability value as the identity estimation probability value.

[0301] The occupation estimation probability value refers to a probability value determined according to the occupation and the coincidence.

[0302] The occupation estimation probability value is generated by weighted calculation of the occupation relevance and the coincidence deviation value, and the occupation estimation probability value is taken as the identity estimation probability value, so as to improve the accuracy of the obtained identity estimation probability value. The specific weighting value for calculating the occupation estimation probability value is set by the operator according to actual needs in advance.

[0303] Based on the same inventive concept, the embodiment of the present application provides a communication clue analysis reminding system, comprising:

[0304] The collection module is configured to collect audio recording information.

[0305] The memory stores a program for implementing the communication clue analysis reminding method.

[0306] The processor loads and executes the program stored in the memory.

[0307] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above functional modules is taken as an example, and in actual application, the above functions can be completed by different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. The specific working process of the system, device and unit described above can refer to the corresponding process in the foregoing method embodiments, which will not be described here.

[0308] The above is only the preferred embodiment of the present application, and the protection scope of the present application is not limited to the above-mentioned embodiments. Any technical solution falling within the concept of the present application shall fall within the protection scope of the present application. It should be noted that, for ordinary skilled persons in the art, some improvements and refinements without departing from the principles of the present application shall also be considered as the protection scope of the present application.

Claims

1. A communication lead analysis alerting method, characterized by, include: S1: Collect audio recording information; S2: Preprocess the audio recording information to extract the speech recording information and environmental recording information; S3: Extract the voiceprint features of the speaker and the recorded text information based on the speech recording information; S4: Input the voiceprint features of the person into a preset voice recognition model to identify and obtain the person's identity information; S5: Combine the person's identity information, the recorded text information, and the environmental recording information to generate a probability value for communication clues; S6: When the probability value of the communication clue is greater than the preset probability benchmark value, generate communication warning information based on the person's identity information and the recorded text information, and output the communication warning information; The method for generating the probability value of the communication clue includes: S51: Based on the person's identity information and the recorded text information, determine the historical text information and the selected text information; S52: Compare the selected text information with the historical text information to determine the text overlap; S53: Combine the environmental recording information with the person's identity information to generate a baseline overlap. S54: Determine whether the text overlap is greater than the baseline overlap; S55: If yes, calculate the difference between the text overlap degree and the benchmark overlap degree, and use it as the overlap deviation value; S56: Combine the overlap deviation value with the person's identity information to generate an identity prediction probability value and use it as the communication clue probability value; S57: If not, output the preset normal probability value and use it as the communication clue probability value; The method for generating the benchmark overlap includes: S531: Determine the identity action range based on the aforementioned person's identity information; S532: Determine the type of the location based on the stated identity and scope of action; S533: Determine the environmental recording type and environmental recording intensity value based on the environmental recording information; S534: Determine whether the environmental recording type belongs to a preset special environmental type; S535: If yes, then determine the environmental prediction type based on the environmental recording type; S536: Combine the environmental prediction type, the range site type and the environmental recording intensity value to generate an environmental prediction overlap degree, and use the environmental prediction overlap degree as the benchmark overlap degree. S537: If not, then determine the estimated overlap of the sites based on the site type of the range, and use the estimated overlap of the sites as the benchmark overlap.

2. The method of claim 1, wherein, The method for determining the historical text information and the selected text information includes: S511: Retrieve the numerical value of the identity and the historical single identity text information based on the aforementioned identity information; S512: Determine whether the value of the identity is only one; S513: If yes, then the historical single identity text information is directly used as the historical text information, and the audio recording text information is used as the selected text information; S514: If not, retrieve simultaneous text information and time-division text information based on the recorded text information; S515: Combine the simultaneous text information and the time-sharing text information to determine the proportion of simultaneous text; S516: Select the person's identity information based on the simultaneous text ratio value to obtain the selected identity information, and use the historical single identity text information corresponding to the selected identity information as the historical text information, and use the audio recording text information corresponding to the selected identity information as the selected text information.

3. The communication clue analysis and alert method according to claim 2, characterized in that, The method for selecting identity information includes: S5161: When the proportion values ​​of the simultaneous text are all greater than the preset benchmark proportion value, retrieve the intensity value of the simultaneous text from the speech recording information based on the simultaneous text information. S5162: Retrieve the single identity strength range based on the aforementioned person identity information; S5163: Combine the simultaneous text intensity value with the single identity intensity range to determine the single identity intensity deviation value; S5164: Combine the single-identity strength deviation value with the simultaneous text strength value to determine the simultaneous text strength reference value; S5165: Select the larger value based on the simultaneous text intensity reference value, and use the corresponding person's identity information as the selected identity information.

4. The communication clue analysis and alert method according to claim 2, characterized in that, The method for selecting identity information further includes: S5166: When the proportion of simultaneous text is not greater than the preset benchmark proportion, the meaning connection value is determined based on the time-sharing text information; S5167: Determine whether the meaning-related value is greater than a preset reference value; S5168: If yes, then all the aforementioned personal identity information shall be used as the selected identity information; S5169: If not, then select the person's identity information based on the simultaneous text proportion value and use it as the selected identity information.

5. The communication clue analysis and alert method according to claim 1, characterized in that, The method for determining the environmental recording type and the environmental recording intensity value includes: S5331: Based on the environmental recording information, retrieve the environmental recording time point, single-time recording frequency value, and single-time recording intensity value; S5332: Combine the single-time recording frequency value with the environmental recording time point to perform curve analysis to form a recording frequency change curve; S5333: Retrieve the frequency change interval time value and the change center frequency value based on the recorded frequency change curve; S5334: Combine the single-time recording intensity value with the environmental recording time point to perform curve analysis to form a recording intensity change curve; S5335: Combine the frequency change interval time value, the change center frequency value, the environmental recording time point and the recording intensity change curve to determine the frequency prediction type, and use the frequency prediction type as the environmental recording type; S5336: Determine the time intensity influence value based on the environmental recording time point; S5337: Determine the selected recording intensity value by combining the recording intensity change curve and the time intensity influence value, and use the selected recording intensity value as the environmental recording intensity value.

6. The communication clue analysis and alert method according to claim 5, characterized in that, The method for generating the frequency prediction type includes: S53351: Determine the frequency similarity by combining the frequency change interval time value and the change center frequency value; S53352: Retrieve the intensity change interval time value and the intensity value at the center of the change based on the recorded intensity change curve; S53353: Determine the interval intensity decay value by combining the intensity change interval time value and the intensity value at the change center; S53354: Determine the estimated echo type by combining the environmental recording time point, the frequency similarity of the interval, and the interval intensity attenuation value; S53355: Determine whether the estimated echo type is greater than the preset echo type benchmark value; S53356: If yes, output the preset echo prediction type and use it as the frequency prediction type; S53357: If not, then the scene prediction type is determined by combining the frequency value of the change center and the intensity value of the change center, and the scene prediction type is used as the frequency prediction type.

7. The communication clue analysis and alert method according to claim 1, characterized in that, The method for generating the environmental prediction overlap includes: S5361: Select the site type within the range based on the environmental prediction type to obtain the selected site type; S5362: Based on the selected site type, determine the estimated location of the type and the benchmark strength value of the type; S5363: Combine the aforementioned type reference intensity value with the aforementioned environmental recording intensity value to determine the environmental intensity deviation value; S5364: Determine the distance estimate based on the environmental intensity deviation value; S5365: Combine the estimated location by type with the estimated distance to determine the estimated range; S5366: Combine the identity action range with the estimated range to determine the range estimated overlap, and use the range estimated overlap as the environment estimated overlap.

8. A communication clue analysis and alerting system, characterized in that, include: The acquisition module is used to acquire audio recording information; A memory storing a program for implementing a communication clue analysis and alerting method as described in any one of claims 1 to 7; The processor loads and executes programs stored in memory.

Images