Method and apparatus for detecting duplicated speech, electronic device, and storage medium

Abstract

This application discloses a method, apparatus, electronic device, and storage medium for detecting copied speech, relating to the field of audio processing. The method includes: acquiring speech data to be detected; performing speech recognition on the speech data to be detected to identify at least two speech segments in the speech data to be detected, the at least two speech segments having the same speech content; performing copied speech detection on the at least two speech segments to obtain a copied speech detection result, wherein the copied speech detection result includes the at least two speech segments being copied speech segments. Compared to brute-force detection of copied segments directly on the original speech data to be detected, this application can improve the efficiency of copied speech detection and reduce the computational load.

Description

Technical Field

[0001] This application relates to the field of audio processing, and more specifically, to a method, apparatus, electronic device, and storage medium for verifying speech identity. Background Technology

[0002] Voice duplication is a common method of forging audio recording evidence; therefore, the detection of copied voice is a crucial aspect of court proceedings in determining the authenticity of audio evidence. Audio signals are random signals; in actual recording, even with identical sound sources, the recorded audio data will not be exactly the same. Therefore, when two identical data segments exist in the voice data, it indicates that these two identical segments must have been artificially copied.

[0003] Currently, common methods for detecting copied speech involve performing comprehensive matching and detection of copied segments across all speech samples. This process is computationally intensive and extremely inefficient for longer audio files. Therefore, improving the efficiency of copied speech detection is a pressing issue that needs to be addressed. Summary of the Invention

[0004] In view of the above problems, embodiments of this application propose a method, apparatus, electronic device and storage medium for detecting copied speech, so as to improve the above problems.

[0005] In a first aspect, this application provides a method for detecting copied speech, the method comprising: acquiring speech data to be detected; performing speech recognition on the speech data to be detected to determine at least two speech segments in the speech data to be detected, the at least two speech segments having the same speech content; performing copied speech detection on the at least two speech segments to obtain a copied speech detection result, wherein the copied speech detection result includes the at least two speech segments being copied speech segments.

[0006] Secondly, this application provides a device for detecting copied speech, the device comprising: a speech data acquisition module for acquiring speech data to be detected; a speech recognition module for performing speech recognition on the speech data to be detected to determine at least two speech segments in the speech data to be detected, the at least two speech segments having the same speech content; and a copied speech detection module for performing copied speech detection on the at least two speech segments to obtain a copied speech detection result, wherein the copied speech detection result includes the at least two speech segments as copied speech segments.

[0007] In some embodiments, the duplicated speech detection module includes: a first feature extraction submodule, configured to extract features from the at least two speech segments to obtain feature sequences corresponding to each of the at least two speech segments; a wallpaper determination submodule, configured to determine the ratio between the feature sequences corresponding to each of the at least two speech segments; and a first duplicated speech segment determination submodule, configured to determine that the at least two speech segments are duplicated speech segments if the ratio is a constant value or the difference between the ratio and the constant value is less than a preset difference.

[0008] In some embodiments, the first determination submodule for copying audio segments includes: a first determination unit, configured to determine, based on the constant value, that the at least two audio segments are obtained by copying and pasting if the ratio is the constant value; and a second determination unit, configured to determine, based on the preset difference, that the at least two audio segments are obtained by copying, scaling, jittering, and pasting if the difference between the ratio and the constant value is less than a preset difference.

[0009] In some embodiments, the constant value includes a first constant value and a second constant value. If the ratio is the constant value, the first determining unit includes: a first determining subunit, configured to determine, if the ratio is the first constant value, that the first speech segment in the at least two speech segments is obtained by copying and pasting the second speech segment in the at least two speech segments; or a second determining subunit, configured to determine, if the ratio is the second constant value, that the first speech segment in the at least two speech segments is obtained by sequentially copying and scaling the second speech segment in the at least two speech segments and then pasting it.

[0010] In some embodiments, the constant value includes a third constant value and a fourth constant value, the preset difference includes a first preset difference and a second preset difference, and the second determining unit includes: a third determining subunit, configured to determine that the first speech segment in the at least two speech segments is obtained by sequentially copying, scaling, and jittering the second speech segment in the at least two speech segments and then pasting it if the difference between the ratio and the third constant value is less than the first preset difference; or a fourth determining subunit, configured to determine that the first speech segment in the at least two speech segments is obtained by sequentially copying, jittering, and scaling the second speech segment in the at least two speech segments and then pasting it if the difference between the ratio and the fourth constant value is less than the second preset difference.

[0011] In some embodiments, the duplicated speech detection module further includes: a second feature extraction submodule, configured to extract features from the at least two speech segments to obtain feature sequences corresponding to each of the at least two speech segments; a difference determination submodule, configured to determine the difference between the feature sequences corresponding to each of the at least two speech segments; and a second duplicated speech segment determination submodule, configured to determine that the at least two speech segments are duplicated speech segments if the difference is a constant value or the difference is within a preset difference range.

[0012] In some embodiments, the second determination submodule for copying speech segments includes: a fifth determination unit, configured to determine, if the difference is a fifth constant value, that the first speech segment among the at least two speech segments is obtained by copying and pasting the second speech segment among the at least two speech segments; or a sixth determination unit, configured to determine, if the difference is within the preset difference range, that the first speech segment among the at least two speech segments is obtained by sequentially copying and jittering the second speech segment among the at least two speech segments before pasting.

[0013] In some embodiments, the duplicated speech detection module further includes: an alignment submodule, configured to align the feature sequences corresponding to each of the at least two speech segments and determine the alignment distance between the feature sequences corresponding to each of the at least two speech segments; a target sequence determination submodule, configured to determine the target sequence corresponding to the feature sequences corresponding to each of the at least two speech segments based on the alignment distance, wherein the target sequence includes at least one of an overlap difference sequence and an overlap ratio sequence; and a duplicated speech detection result determination submodule, configured to perform duplicated speech detection on the target sequence if the target sequence meets the alignment condition, and obtain the duplicated speech detection result.

[0014] In some embodiments, the device for detecting copied speech further includes: a speech duration determination module, configured to determine the timestamp of the copied speech segment in the speech data to be detected, and determine the speech duration of the copied speech segment based on the timestamp; and a target copied speech segment determination module, configured to determine a target copied speech segment from the copied speech segment whose speech duration is greater than a preset duration.

[0015] Thirdly, embodiments of this application provide an electronic device, including: a processor; and a memory, wherein computer-readable instructions are stored in the memory, and when the computer-readable instructions are executed by the processor, the method for detecting copied speech as described above is implemented.

[0016] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-readable instructions thereon, which, when executed by a processor, implement the method for detecting copied speech as described above.

[0017] In this application, speech recognition is performed on the speech data to be detected to extract at least two speech segments with the same speech content. Then, duplicate speech detection is performed based on these at least two speech segments to determine the duplicate speech detection result corresponding to the speech data to be detected. Compared to brute-force detection of duplicate segments directly applied to the original speech data to be detected, this application improves the efficiency of duplicate speech detection and reduces the computational load.

[0018] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit the invention. Attached Figure Description

[0019] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application. It is obvious that the drawings described below are merely some embodiments of this application, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort.

[0020] Figure 1 This is a schematic flowchart illustrating a method for detecting copied speech according to an embodiment of this application.

[0021] Figure 2 This is a flowchart illustrating a method for detecting copied speech according to another embodiment of this application.

[0022] Figure 3 This is a schematic flowchart illustrating the specific steps of step 250 according to an embodiment of this application.

[0023] Figure 4 This is a schematic flowchart illustrating a method for detecting copied speech in another embodiment of this application.

[0024] Figure 5 This is a schematic flowchart illustrating a method for detecting copied speech in another embodiment of this application.

[0025] Figure 6 This is a schematic flowchart illustrating a method for detecting copied speech in another embodiment of this application.

[0026] Figure 7 This is a block diagram of a device for detecting copied speech according to an embodiment of this application.

[0027] Figure 8 This is a hardware structure diagram of an electronic device according to an embodiment of this application.

[0028] The accompanying drawings have illustrated specific embodiments of the present invention, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the inventive concept in any way, but rather to illustrate the concept of the invention to those skilled in the art through specific embodiments. Detailed Implementation

[0029] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided to make this application more comprehensive and complete, and to fully convey the concept of the exemplary embodiments to those skilled in the art.

[0030] Furthermore, the described features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. Numerous specific details are provided in the following description to give a thorough understanding of embodiments of this application. However, those skilled in the art will recognize that the technical solutions of this application can be practiced without one or more of the specific details, or other methods, components, apparatuses, steps, etc., can be employed. In other instances, well-known methods, apparatuses, implementations, or operations are not shown or described in detail to avoid obscuring various aspects of this application.

[0031] Please see Figure 1 , Figure 1 This application illustrates a method for detecting copied speech according to an embodiment of the present application. In a specific embodiment, this method for detecting copied speech can be applied to, for example... Figure 7 The illustrated device 600 for detecting copied speech and the electronic device 700 equipped with the device 600 for detecting copied speech are shown. Figure 8 The specific process of this embodiment will be described below. Of course, it is understood that this method can be executed by an electronic device with computing capabilities, such as a desktop computer, laptop computer, in-vehicle terminal, or in-vehicle infotainment screen. This method can also be interactively executed by a processing system including a server and a terminal. The following will focus on... Figure 1 The process shown will be described in detail. The method for detecting copied speech may specifically include the following steps: Step 110: Obtain the voice data to be detected.

[0032] The voice data to be detected refers to the voice data that needs to be copied. This voice data to be detected may be voice data submitted as evidence in court.

[0033] Step 120: Perform speech recognition on the speech data to be detected to determine at least two speech segments in the speech data to be detected, wherein the at least two speech segments have the same speech content.

[0034] Automatic Speech Recognition (ASR) is a field that studies speech and uses speech signal processing and pattern recognition to enable machines to automatically recognize and understand human speech and convert the speech signal into corresponding text.

[0035] The audio content can be either text or phoneme content; no specific limitation is made here.

[0036] One approach is to use a speech recognition model to perform speech recognition on the speech data to be detected. This model identifies the text content corresponding to an audio clip. Optionally, the language to be output by the speech recognition model can be specified. For example, if the input audio is in English, the output text can be either Chinese or English. The language of the output text can be set according to actual needs; no specific limitations are imposed here.

[0037] Optionally, to ensure the accuracy of speech recognition results, the speech recognition model needs to be trained in advance. Specifically, a sample set is pre-constructed, which includes multiple sample speech data and the corresponding speech content for each sample speech data. The speech content for each sample speech data can be text content. During training, the sample speech data is input into the speech recognition model for speech recognition, obtaining the recognition result for that sample speech data. This recognition result indicates the text content of the sample speech data. Then, based on the speech content and recognition result of each sample speech data, the loss function is calculated. If the loss value does not converge, the parameters of the speech recognition model are adjusted in reverse, and the adjusted speech recognition model outputs the sample recognition result again for the same sample speech data, and the loss function is calculated again until the loss value converges. This process is repeated for each sample speech data. When the training termination condition is met, the training of the speech recognition model ends. Afterwards, the speech recognition model is used online for object recognition, ensuring the accuracy of speech recognition for the target speech data.

[0038] One approach is to perform voice activity detection (VAD) on the speech data to be recognized before performing speech recognition. This distinguishes between active and inactive speech segments within the speech data, and then performs speech recognition based on the active speech. VAD, also known as endpoint detection, distinguishes between speech segments and non-speech segments (also called silence segments) in an audio file, removing silence segments and retaining only the speech segments. Therefore, after performing VAD on the speech data, non-speech segments can be filtered out, leaving only the active speech segments. During speech recognition, only the active speech segments need to be recognized, eliminating the need to focus on non-speech segments and reducing their interference with speech recognition, thus improving the accuracy of speech recognition for the speech data.

[0039] As another approach, noise reduction processing can be performed on the speech data to be detected before speech recognition, thereby avoiding the impact of noise data in the speech data on the speech recognition of the speech data to be detected, and thus reducing the accuracy of speech recognition of the speech data to be detected.

[0040] Step 130: Perform duplicate speech detection on the at least two speech segments to obtain duplicate speech detection results, wherein the duplicate speech detection results include the at least two speech segments being duplicate speech segments.

[0041] Voice duplication is a common method of forging audio recording evidence. Since audio signals are random signals, even if the sound source is exactly the same (i.e., the same person), the recorded voice data cannot be exactly the same. Therefore, voice duplication detection of voice data submitted as evidence is an important part of the court's verification of the authenticity of the recording.

[0042] Optionally, the copied speech segment can be a speech fragment obtained by copying and pasting from the speech data to be detected. Specifically, this includes speech fragments obtained by directly copying and pasting, as well as speech fragments obtained by copying, scaling proportionally, and then pasting.

[0043] As one approach, since at least two speech segments with identical speech content are not necessarily duplicate speech segments, duplicate segment detection is required to determine whether at least two speech segments with identical speech content are duplicate speech segments. Optionally, the spectrograms of at least two speech segments are compared to determine whether the spectrograms of at least two speech segments are completely identical. If the spectrograms of at least two speech segments are completely identical, the at least two speech segments can be determined to be duplicate speech segments; or if the spectrograms of at least two speech segments show proportional scaling, the at least two speech segments can be determined to be duplicate speech segments; or if the similarity between the spectrograms of at least two speech segments is greater than a similarity threshold, the at least two speech segments can be determined to be duplicate speech segments.

[0044] As an alternative approach, at least two speech segments can be sampled first to determine the audio features corresponding to each sampling point in the at least two speech segments. The audio features of each sampling point in the at least two speech segments are then compared to identify consecutive sampling points with the same audio features. The audio segment corresponding to these consecutive sampling points with the same audio features is the replicated speech segment. The audio features of each sampling point can be the amplitude, spectrum, cepstral, formants, pitch, reflection coefficient, etc., of the speech segment at each sampling point, and are not specifically limited here.

[0045] As another approach, when the number of speech segments is greater than two, the copy detection result for these speech segments also includes at least two speech segments being copied, and the remaining speech segments not being copied from other speech segments. Optionally, the copy detection result also includes all speech segments not being copied from each other.

[0046] In the embodiments of this application, speech recognition is performed on the speech data to be detected to extract at least two speech segments with the same speech content. Then, duplicate speech detection can be performed based on these at least two speech segments to determine the duplicate speech detection result corresponding to the speech data to be detected. Compared to brute-force detection of duplicate segments directly applied to the original speech data to be detected, this application improves the efficiency of duplicate speech detection and reduces the computational load.

[0047] Please see Figure 2 , Figure 2 This application illustrates a method for detecting copied speech according to an embodiment of the present application. In a specific embodiment, this method for detecting copied speech can be applied to, for example... Figure 7 The illustrated device 600 for detecting copied speech and the electronic device 700 equipped with the device 600 for detecting copied speech are shown. Figure 8The specific process of this embodiment will be described below. Of course, it is understood that this method can be executed by an electronic device with computing capabilities, such as a desktop computer, laptop computer, in-vehicle terminal, or in-vehicle infotainment screen. This method can also be interactively executed by a processing system including a server and a terminal. The following will focus on... Figure 2 The process shown will be described in detail. The method for detecting copied speech may specifically include the following steps: Step 210: Obtain the voice data to be detected.

[0048] Step 220: Perform speech recognition on the speech data to be detected to determine at least two speech segments in the speech data to be detected, wherein the at least two speech segments have the same speech content.

[0049] For a detailed description of steps 210 and 220, please refer to steps 110 and 120, which will not be repeated here.

[0050] Step 230: Extract features from the at least two speech segments to obtain the feature sequences corresponding to each of the at least two speech segments.

[0051] As one approach, the feature sequence can be a sequence of features such as fundamental frequency, energy, and zero-crossing rate corresponding to at least two speech segments. Optionally, the features can also be formant trends, Mel-frequency cepstral coefficients, harmonics, etc. Optionally, the fundamental frequency, energy, and zero-crossing rate of each frame of the speech signal in at least two speech segments can be automatically determined using speech signal processing algorithms, such as autocorrelation, cepstral method, and linear prediction (LPC). Then, the feature sequences corresponding to each of the at least two speech segments can be determined based on the obtained features such as fundamental frequency, energy, and zero-crossing rate of each frame of the speech signal in at least two speech segments.

[0052] Alternatively, the speech segments of at least two people can be sampled separately to determine the feature values ​​corresponding to each sampling point of the speech segments of at least two people. Then, based on the order of the sampling points of the speech segments of at least two people and the feature values ​​corresponding to each sampling point, the feature sequences corresponding to each of the at least two speech segments can be determined. Optionally, the number of sampling points and the sampling frequency of the at least two speech segments can be set according to actual needs, and are not specifically limited here.

[0053] Step 240: Determine the ratio between the feature sequences corresponding to each of the at least two speech segments.

[0054] As one approach, the feature values ​​corresponding to each sampling point are proportional, and their proportionality coefficients are the same. Based on this principle, it is possible to determine whether at least two speech segments are duplicated speech segments by calculating the ratio between the feature sequences corresponding to each of the at least two speech segments.

[0055] Optionally, the feature values ​​corresponding to the same sampling point in the feature sequences of at least two speech segments can be divided, and it can be determined whether the ratios of each sampling point are the same. Optionally, a threshold for the number of sampling point groups that meets a preset condition can be preset, and the number of sampling point groups that meet the preset condition in the feature sequences of at least two speech segments can be counted. When the number of groups is greater than the threshold, the ratio between the sampling points can be used as the ratio between the feature sequences corresponding to each of the at least two speech segments. Here, the preset condition can be that the ratios of the feature values ​​of different groups of sampling points are the same. Optionally, the preset condition can also be that the ratios of the feature values ​​of different groups of sampling points between any two feature sequences corresponding to at least two speech segments are the same.

[0056] Optionally, any feature sequence can be selected from the feature sequences corresponding to at least two speech segments. The feature value of the first sampling point in the feature sequence is divided by the feature values ​​of each sampling point in the feature sequences corresponding to other speech segments to determine the target sampling point in the other feature sequences that has the same feature value or is proportional to the first sampling point. Then, the two feature sequences are aligned based on the target sampling point, and the ratio between the two feature sequences is determined.

[0057] As another approach, before determining the ratio between the feature sequences corresponding to at least two speech segments, the sampling points are filtered based on the feature values ​​corresponding to each sampling point. This avoids situations where the ratio between the feature sequences corresponding to at least two speech segments cannot be determined. Optionally, the sampling points can be filtered by determining whether there are any sampling points with feature values ​​less than a feature value threshold. If so, the sampling point is filtered out. Since only the integer part of the feature value of the sampling point is calculated when calculating the ratio between each sampling point, to avoid calculating the ratio of feature values ​​with an integer part of 0, the feature value threshold can be set to 1. Other feature value thresholds can also be set, without specific limitations here.

[0058] Step 250: If the ratio is a constant value or the difference between the ratio and the constant value is less than a preset difference, then the at least two speech segments are determined to be copied speech segments.

[0059] In one approach, if the ratio between the corresponding feature sequences of mutually duplicated speech segments is a constant or tends to be a constant, then the ratio calculated in step 240 can be used to determine whether at least two speech segments are duplicated speech segments.

[0060] In some embodiments, such as Figure 3 As shown, step 250 includes: Step 251: If the ratio is the constant value, then based on the constant value, it is determined that the at least two speech segments were obtained by copying and pasting.

[0061] In some embodiments, the constant value includes a first constant value and a second constant value, and step 251 includes: if the ratio is the first constant value, then determining that the first speech segment of the at least two speech segments is obtained by copying and pasting the second speech segment of the at least two speech segments; if the ratio is the second constant value, then determining that the first speech segment of the at least two speech segments is obtained by sequentially copying and scaling the second speech segment of the at least two speech segments and then pasting it.

[0062] As one approach, if the first speech segment is obtained by directly copying and pasting the second speech segment, the feature sequence corresponding to the first speech segment is exactly the same as the feature sequence corresponding to the second speech segment. Then, the corresponding first ratio is the first constant value, where the first constant value can be 1.

[0063] One approach is to copy and scale the second speech segment, which alters its feature values. The change in feature values ​​is negatively correlated with the scaling ratio; that is, the larger the scaling ratio, the greater the difference between the corresponding feature values ​​and the original feature values ​​of the second speech segment. However, between corresponding sampling points of the first and second speech segments, the corresponding feature values ​​are proportional, and the proportion is the same for all corresponding sampling points. Therefore, when the ratio is determined to be a second constant value, it can be determined that the first speech segment among at least two speech segments is obtained by sequentially copying and scaling the second speech segment among at least two speech segments and then pasting it. Optionally, the second constant value differs from the first constant value and is related to the scaling ratio. For example, if the second speech segment is copied, scaled by 50%, and then pasted to obtain the first speech segment, the ratio between the feature sequences of the first and second speech segments is 2 or 1 / 2. Optionally, the second constant value can also be other values; this is only an example and not a specific limitation.

[0064] Step 252: If the difference between the ratio and the constant value is less than a preset difference, then the at least two speech segments are determined to have been obtained by copying, scaling, jittering, and pasting based on the preset difference.

[0065] In some embodiments, the constant value includes a third constant value and a fourth constant value, and the preset difference includes a first preset difference and a second preset difference. Step 252 includes: if the difference between the ratio and the third constant value is less than the first preset difference, then it is determined that the first speech segment among the at least two speech segments is obtained by sequentially copying, scaling, and jittering the second speech segment among the at least two speech segments and then pasting it; if the difference between the ratio and the fourth constant value is less than the second preset difference, then it is determined that the first speech segment among the at least two speech segments is obtained by sequentially copying, jittering, and scaling the second speech segment among the at least two speech segments and then pasting it.

[0066] Dithering achieves the same auditory effect as using more bits with fewer bits by adding noise to a frame of speech data. For example, to make 16-bit speech data sound like 20-bit speech data, the simplest method is to remove the last 4 bits of the 20-bit speech data. However, this reduces the detail in the speech data. Because the human ear is particularly sensitive to periodic sounds, prominent frequencies in the spectrum, and prominent frequencies that shift over time, discarding the last bits of the speech data results in low correlation with the original speech data, leading to significant incompleteness. The human ear is very sensitive to the abruptness caused by incomplete speech data. Dithering, on the other hand, adds noise to bits 17-20 of the speech data. This noise may carry over and change the information in the 16th bit. Then, the last 4 bits are removed; this process is called redithering. This linearly reflects the data from the last 4 bits of the original speech data onto the 16th bit. Because the human ear has the ability to easily separate noise from musical sounds, although noise was added, more details of the speech data were actually heard, and there was no significant difference in hearing between the original speech data and the jittered speech data. Optionally, the added noise can be white noise or other types of noise; no specific limitation is made here.

[0067] As one approach, jittering of the second speech segment causes the feature values ​​of each sampling point to differ from those of the original speech data. This results in a change in the ratio between the feature sequence of the jittered speech data and the feature sequence of the original speech data. However, this ratio tends to a constant value, meaning that the absolute value of the difference between this ratio and the constant value is less than the difference threshold.

[0068] Based on the above description, the difference between the ratio of the feature sequences corresponding to the first and second speech segments and the third constant value can be used to determine whether the first speech segment in at least two speech segments was obtained by sequentially copying, scaling, and jittering the second speech segment in at least two speech segments before pasting. The third constant value and the second constant value can be the same or different, and can be determined according to actual needs; no specific limitation is made here.

[0069] As one approach, different orders of jitter and scaling after copying the second speech segment will result in different first speech segments, i.e., different feature sequences corresponding to the first speech segments, which in turn will lead to different ratios between the feature sequences of the first and second speech segments. Correspondingly, different constant values ​​can be determined to distinguish the order of jitter and scaling after copying the second speech segment; that is, the fourth constant value differs from the third constant value. Optionally, the first preset difference and the second preset difference can be the same or different, depending on actual needs, and are not limited here.

[0070] In this embodiment, duplicate speech is detected by dividing the feature sequences corresponding to speech segments with the same speech content by the ratio. The duplicate speech segment and the method of obtaining the duplicate speech segment are determined by comparing the ratio with a constant value or a preset difference, thereby improving the accuracy and efficiency of duplicate speech detection.

[0071] Please see Figure 4 , Figure 4 This application illustrates a method for detecting copied speech according to an embodiment of the present application. In a specific embodiment, this method for detecting copied speech can be applied to, for example... Figure 7 The illustrated device 600 for detecting copied speech and the electronic device 700 equipped with the device 600 for detecting copied speech are shown. Figure 8 The specific process of this embodiment will be described below. Of course, it is understood that this method can be executed by an electronic device with computing capabilities, such as a desktop computer, laptop computer, in-vehicle terminal, or in-vehicle infotainment screen. This method can also be interactively executed by a processing system including a server and a terminal. The following will focus on... Figure 4 The process shown will be described in detail. The method for detecting copied speech may specifically include the following steps: Step 310: Obtain the voice data to be detected.

[0072] Step 320: Perform speech recognition on the speech data to be detected to determine at least two speech segments in the speech data to be detected, wherein the at least two speech segments have the same speech content.

[0073] Step 330: Extract features from the at least two speech segments to obtain the feature sequences corresponding to each of the at least two speech segments.

[0074] For a detailed description of steps 310-330, please refer to steps 110-120 and 230, which will not be repeated here.

[0075] Step 340: Determine the difference between the feature sequences corresponding to each of the at least two speech segments.

[0076] As one approach, if two speech segments are duplicated speech segments, the feature values ​​corresponding to each sampling point in their corresponding feature sequences are the same. Based on this principle, it is possible to determine whether at least two speech segments are duplicated speech segments by calculating the difference between the feature sequences corresponding to each of the at least two speech segments.

[0077] Step 350: If the difference is a constant value or the difference is within a preset difference range, then the at least two speech segments are determined to be copied speech segments.

[0078] In one approach, if the difference between the corresponding feature sequences of mutually duplicated speech segments is a constant or tends to be a constant, then the difference calculated in step 340 can be used to determine whether at least two speech segments are duplicated speech segments.

[0079] In some embodiments, step 350 includes: if the difference is a fifth constant value, then determining that the first speech segment among the at least two speech segments is obtained by copying and pasting the second speech segment among the at least two speech segments; if the difference is within the preset difference range, then determining that the first speech segment among the at least two speech segments is obtained by sequentially copying and jittering the second speech segment among the at least two speech segments and then pasting it.

[0080] As one approach, if the first speech segment is obtained by directly copying and pasting the second speech segment, the feature sequence corresponding to the first speech segment is exactly the same as the feature sequence corresponding to the second speech segment. The difference between them is the fifth constant value, which can be 0.

[0081] As one approach, copying the speech data and then performing jitter processing will correspondingly change the feature values ​​of each sampling point of the speech data. However, jitter processing alone will not cause a significant change in the feature values ​​of each sampling point; the changes will only occur within a certain range. Therefore, the determination can be made based on whether the difference between the feature sequences corresponding to the first speech segment and the feature sequences corresponding to the second speech segment is within a preset difference range. Specifically, this involves determining whether the first speech segment out of at least two speech segments was obtained by sequentially copying and jittering the second speech segment out of at least two speech segments and then pasting it. The preset difference range can be [-1, 1] or other difference ranges, without specific limitations here.

[0082] In this embodiment, duplicate speech is detected by subtracting the feature sequences corresponding to speech segments with the same speech content. The duplicate speech segment and the method of obtaining the duplicate speech segment are determined by comparing the difference with a constant value or a preset difference range, thereby improving the accuracy and efficiency of duplicate speech detection.

[0083] Please see Figure 5 , Figure 5 This application illustrates a method for detecting copied speech according to an embodiment of the present application. In a specific embodiment, this method for detecting copied speech can be applied to, for example... Figure 7 The illustrated device 600 for detecting copied speech and the electronic device 700 equipped with the device 600 for detecting copied speech are shown. Figure 8 The specific process of this embodiment will be described below. Of course, it is understood that this method can be executed by an electronic device with computing capabilities, such as a desktop computer, laptop computer, in-vehicle terminal, or in-vehicle infotainment screen. This method can also be interactively executed by a processing system including a server and a terminal. The following will focus on... Figure 5 The process shown will be described in detail. The method for detecting copied speech may specifically include the following steps: Step 410: Obtain the speech data to be detected.

[0084] Step 420: Perform speech recognition on the speech data to be detected to determine at least two speech segments in the speech data to be detected, wherein the at least two speech segments have the same speech content.

[0085] For a detailed description of steps 410-420, please refer to steps 110-120, which will not be repeated here.

[0086] Step 430: Align the feature sequences corresponding to each of the at least two speech segments to determine the alignment distance between the feature sequences corresponding to each of the at least two speech segments.

[0087] As one approach, for at least two speech segments with identical speech content, the sampled data may not be aligned. That is, even if at least two speech segments with identical content are artificially copied, their speech lengths may not be the same, and the positions of their corresponding sampling points may not correspond one-to-one. This can lead to the ratio or difference between the feature sequences corresponding to the at least two copied speech segments not meeting the requirements, resulting in an incorrect detection of copied speech in the target speech data. Therefore, before performing copied speech detection, at least two speech segments need to be aligned. Optionally, alignment can be performed using maximum correlation.

[0088] Alignment distance refers to the distance that at least one speech segment needs to move relative to any other speech segment when aligning at least two speech segments. Optionally, this alignment distance can be the number of sampling points that the feature sequence corresponding to at least one speech segment needs to move when aligning with the feature sequence corresponding to any other speech segment; alternatively, this alignment distance can also be the duration of the speech segment that at least one speech segment moves relative to any other speech segment after aligning at least two speech segments according to their corresponding feature sequences. This can be set according to actual needs and is not specifically limited here.

[0089] As one approach, the alignment of feature sequences corresponding to at least two speech segments can be performed using a correlation function; alternatively, the formula can be used: in, For any one of at least two speech segments, the feature sequence is... For at least two speech segments, the feature sequence corresponding to the other speech segment. For the correlation function, through the correlation function Determine the distance m that the feature sequences corresponding to at least two speech segments need to move after alignment, where n is the index of the sampling point in the feature sequences corresponding to at least two speech segments.

[0090] Step 440: Based on the alignment distance, determine the target sequence corresponding to the feature sequence of each of the at least two speech segments, wherein the target sequence includes at least one of an overlap difference sequence and an overlap ratio sequence.

[0091] One approach is to determine the overlapping sampling points based on the feature sequences corresponding to at least two aligned speech segments after determining the alignment distance, and then determine the target sequence based on these overlapping sampling points. The target sequence is also known as the overlapping sequence.

[0092] Optionally, the overlap difference sequence or overlap ratio sequence refers to the sequence obtained by subtracting the overlapping sequences after alignment of the feature sequences corresponding to at least two speech segments; the overlap ratio sequence refers to the sequence obtained by dividing the overlapping sequences after alignment of the feature sequences corresponding to at least two speech segments. This overlap difference sequence or overlap ratio sequence can be used for duplicate speech detection. The overlap ratio sequence can be determined according to... Figure 2 The illustrated embodiment performs duplicate speech detection, and the overlapping difference sequence can be based on... Figure 4 The illustrated embodiment performs duplicate speech detection.

[0093] Optionally, the overlapping difference sequences can be determined according to the formula. It is confirmed that, among them, It is an overlapping difference sequence. The length of the overlapping sequence (sum of sampling points). It is a sequence aligned with the feature sequences corresponding to any one of at least two speech segments. A sequence aligned with the feature sequences corresponding to another speech segment out of at least two speech segments is, understandably, and This is the aligned state, where k is the sampling point in the overlapping difference sequence. For each sampling point in the overlapping difference sequence The corresponding sampling points, For each sampling point in the overlapping difference sequence The corresponding sampling points. Optional, ; Where m is the alignment distance. The sampling points of the overlapping difference sequence correspond to the The order of sampling points in the original feature sequence. The sampling points of the overlapping difference sequence correspond to the The order of sampling points in the original feature sequence.

[0094] Optionally, the overlap ratio sequence can be determined according to the formula. It is confirmed that, among them, It is an overlap ratio sequence. The length of the overlapping sequence (sum of sampling points). It is a sequence aligned with the feature sequences corresponding to any one of at least two speech segments. A sequence aligned with the feature sequences corresponding to another speech segment out of at least two speech segments is, understandably, and This is the aligned state, where k is the sampling point in the overlapping difference sequence. For each sampling point in the overlapping difference sequence The corresponding sampling points, For each sampling point in the overlapping difference sequence The corresponding sampling points. Optional, ; Where m is the alignment distance. The sampling points of the overlapping difference sequence correspond to the The order of sampling points in the original feature sequence. The sampling points of the overlapping difference sequence correspond to the The order of sampling points in the original feature sequence.

[0095] As one approach, after determining the alignment distance, when determining the target sequence, the alignment distance is a fixed value, and the target sequence can be determined directly based on the alignment distance.

[0096] Step 450: If the target sequence meets the alignment condition, then perform duplicate speech detection on the target sequence to obtain the duplicate speech detection result.

[0097] One approach is to determine whether the difference between the feature values ​​of each overlapping sampling point in the overlapping difference sequence is within a preset range. For example, if the difference between the feature values ​​of each overlapping sampling point is between [-1, 1], it can be determined that the overlapping difference sequence meets the alignment condition. Then, based on... Figure 4 The illustrated embodiment performs duplicate speech detection on overlapping differential sequences to determine duplicate speech segments and the process of acquiring duplicate speech segments.

[0098] As another approach, the alignment condition can also be whether the ratio between the feature values ​​of each overlapping sampling point in the overlap ratio sequence is within a preset range. Optionally, if there are overlapping sampling points with a feature value of 0 among the feature values ​​of overlapping sampling points, then the ratio of that overlapping sampling point can be determined to be 0. Optionally, to avoid situations where the ratio between the feature values ​​of each overlapping sampling point in the overlap ratio sequence is not within the preset range due to the possible simultaneous existence of dithering and scaling during the copying process of at least two speech segments, a target overlapping sampling point with a feature value greater than or equal to a feature value threshold can be determined among the overlapping sampling points in the overlap ratio sequence. Whether the target sequence meets the alignment condition is determined based on whether the ratio of the feature values ​​of the target overlapping sampling point is within the preset range. When the overlap ratio sequence meets the alignment condition, according to... Figure 2 The illustrated embodiment performs duplicate speech detection on the overlap ratio sequence to determine the duplicate speech segment and the process of acquiring the duplicate speech segment.

[0099] As another approach, it is possible to determine whether a speech segment is a copy while aligning the feature sequences corresponding to at least two speech segments. Optionally, after determining the feature sequences corresponding to at least two speech segments, any one feature sequence is selected as a reference feature sequence. Using the first sampling point of the reference feature sequence as a benchmark, the sampling points of other feature sequences are sequentially aligned with the first to last sampling points of the reference feature sequence to determine overlapping sampling points, and thus determine the overlapping difference sequence or overlapping ratio sequence. Optionally, aligning the sampling points of other feature sequences with the first to last sampling points of the reference feature sequence can be achieved by subtracting or dividing the feature values ​​of each sampling point of the other feature sequences from the first to last sampling points of the reference feature sequence, and determining whether the difference or ratio between the feature values ​​of each sampling point meets the alignment condition. When the difference or ratio between the feature values ​​of each sampling point meets the alignment condition, it also meets the condition for copy speech detection, thereby identifying the copy speech segment among the at least two speech segments.

[0100] In this embodiment, the alignment distance is determined by aligning the feature sequences corresponding to speech segments with the same speech content. Then, the target feature sequence is determined in the feature sequences corresponding to speech segments with the same speech content based on the alignment distance. This enables duplicate speech detection when the target feature sequence meets the alignment conditions, ensuring that the speech data to be detected with duplicate speech segments is not missed, and improving the accuracy of duplicate speech detection.

[0101] Please see Figure 6 , Figure 6 This application illustrates a method for detecting copied speech according to an embodiment of the present application. In a specific embodiment, this method for detecting copied speech can be applied to, for example... Figure 7 The illustrated device 600 for detecting copied speech and the electronic device 700 equipped with the device 600 for detecting copied speech are shown. Figure 8 The specific process of this embodiment will be described below. Of course, it is understood that this method can be executed by an electronic device with computing capabilities, such as a desktop computer, laptop computer, in-vehicle terminal, or in-vehicle infotainment screen. This method can also be interactively executed by a processing system including a server and a terminal. The following will focus on... Figure 6 The process shown will be described in detail. The vehicle control method may specifically include the following steps: Step 510: Obtain the speech data to be detected; Step 520: Perform speech recognition on the speech data to be detected to determine at least two speech segments in the speech data to be detected, wherein the at least two speech segments have the same speech content; Step 530: Perform duplicate speech detection on the at least two speech segments to obtain duplicate speech detection results, wherein the duplicate speech detection results include the at least two speech segments being duplicate speech segments.

[0102] For a detailed description of steps 510-530, please refer to steps 110-130, which will not be repeated here.

[0103] Step 540: Determine the timestamp of the copied speech segment in the speech data to be detected, and determine the speech duration of the copied speech segment based on the timestamp.

[0104] One approach is to include the start and end timestamps of the copied speech segment in the speech data to be detected, and determine the duration of the copied speech segment based on the start and end timestamps. Optionally, the timestamp of the copied speech segment can be determined by simultaneously identifying the time information of at least two speech segments with the same speech content in the speech data to be detected during speech recognition, thus facilitating the subsequent determination of the timestamp of the copied speech segment in the speech data to be detected.

[0105] Step 550: Determine a target copied speech segment from the copied speech segments whose speech duration is greater than a preset duration.

[0106] As one approach to ensure the validity of copied speech segments, a preset speech duration can be established. A speech segment is considered valid only if its duration exceeds this preset duration, guaranteeing that it contains valid information. This is because, to prove the validity of the copied speech segment, the duration of each determined copied speech segment is compared with the preset duration. This comparison identifies target copied speech segments with durations exceeding the preset duration, thus ensuring the effectiveness of the copied speech detection.

[0107] In this embodiment, the duration of each copied speech segment is determined based on the timestamp of the copied speech segment in the speech data to be detected. This allows the target copied speech segment to be determined based on the speech duration, thereby ensuring the effectiveness of copied speech detection and improving its accuracy.

[0108] Figure 7 This is a block diagram of a device for detecting copied speech according to an embodiment of this application, as shown below. Figure 7 As shown, the device 600 for detecting copied speech includes: a speech data acquisition module 610, a speech recognition module 620, and a copied speech detection module 630.

[0109] The speech data acquisition module 610 is used to acquire speech data to be detected; the speech recognition module 620 is used to perform speech recognition on the speech data to be detected and determine at least two speech segments in the speech data to be detected, wherein the at least two speech segments have the same speech content; the duplicate speech detection module 630 is used to perform duplicate speech detection on the at least two speech segments and obtain a duplicate speech detection result, wherein the duplicate speech detection result includes whether the at least two speech segments are duplicate speech segments or whether the at least two speech segments are not duplicate speech segments.

[0110] In some embodiments, the duplicate speech detection module 630 includes: a first feature extraction submodule, configured to extract features from the at least two speech segments to obtain feature sequences corresponding to each of the at least two speech segments; a wallpaper determination submodule, configured to determine the ratio between the feature sequences corresponding to each of the at least two speech segments; and a first duplicate speech segment determination submodule, configured to determine that the at least two speech segments are duplicate speech segments if the ratio is a constant value or the difference between the ratio and the constant value is less than a preset difference.

[0111] In some embodiments, the first determination submodule for copying audio segments includes: a first determination unit, configured to determine, based on the constant value, that the at least two audio segments are obtained by copying and pasting if the ratio is the constant value; and a second determination unit, configured to determine, based on the preset difference, that the at least two audio segments are obtained by copying, scaling, jittering, and pasting if the difference between the ratio and the constant value is less than a preset difference.

[0112] In some embodiments, the constant value includes a first constant value and a second constant value. If the ratio is the constant value, the first determining unit includes: a first determining subunit, configured to determine, if the ratio is the first constant value, that the first speech segment in the at least two speech segments is obtained by copying and pasting the second speech segment in the at least two speech segments; or a second determining subunit, configured to determine, if the ratio is the second constant value, that the first speech segment in the at least two speech segments is obtained by sequentially copying and scaling the second speech segment in the at least two speech segments and then pasting it.

[0113] In some embodiments, the constant value includes a third constant value and a fourth constant value, the preset difference includes a first preset difference and a second preset difference, and the second determining unit includes: a third determining subunit, configured to determine that the first speech segment in the at least two speech segments is obtained by sequentially copying, scaling, and dithering the second speech segment in the at least two speech segments and then pasting it if the difference between the ratio and the third constant value is less than the first preset difference; or a fourth determining subunit, configured to determine that the first speech segment in the at least two speech segments is obtained by sequentially copying, dithering, and scaling the second speech segment in the at least two speech segments and then pasting it if the difference between the ratio and the fourth constant value is less than the second preset difference.

[0114] In some embodiments, the duplicated speech detection module 630 further includes: a second feature extraction submodule, configured to extract features from the at least two speech segments to obtain feature sequences corresponding to each of the at least two speech segments; a difference determination submodule, configured to determine the difference between the feature sequences corresponding to each of the at least two speech segments; and a second duplicated speech segment determination submodule, configured to determine that the at least two speech segments are duplicated speech segments if the difference is a constant value or the difference is within a preset difference range.

[0115] In some embodiments, the second determination submodule for copying speech segments includes: a fifth determination unit, configured to determine, if the difference is a fifth constant value, that the first speech segment among the at least two speech segments is obtained by copying and pasting the second speech segment among the at least two speech segments; or a sixth determination unit, configured to determine, if the difference is within the preset difference range, that the first speech segment among the at least two speech segments is obtained by sequentially copying and jittering the second speech segment among the at least two speech segments before pasting.

[0116] In some embodiments, the duplicate speech detection module 630 further includes: an alignment submodule, configured to align the feature sequences corresponding to each of the at least two speech segments and determine the alignment distance between the feature sequences corresponding to each of the at least two speech segments; a target sequence determination submodule, configured to determine the target sequence corresponding to the feature sequences corresponding to each of the at least two speech segments based on the alignment distance, wherein the target sequence includes at least one of an overlap difference sequence and an overlap ratio sequence; and a duplicate speech detection result determination submodule, configured to perform duplicate speech detection on the target sequence if the target sequence meets the alignment condition, and obtain the duplicate speech detection result.

[0117] In some embodiments, the device 600 for detecting copied speech further includes: a speech duration determination module, configured to determine the timestamp of the copied speech segment in the speech data to be detected, and determine the speech duration of the copied speech segment based on the timestamp; and a target copied speech segment determination module, configured to determine a target copied speech segment from the copied speech segment whose speech duration is greater than a preset duration.

[0118] According to one aspect of the embodiments of this application, an electronic device is also provided, such as... Figure 8 As shown, the electronic device 700 includes a processor 710 and one or more memories 720. The one or more memories 720 are used to store program instructions executed by the processor 710. When the processor 710 executes the program instructions, it implements the above-described method for detecting copied speech.

[0119] Furthermore, the processor 710 may include one or more processing cores. The processor 710 runs or executes instructions, programs, code sets, or instruction sets stored in the memory 720, and retrieves data stored in the memory 720. Optionally, the processor 710 may be implemented using at least one hardware form selected from Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), and Programmable Logic Array (PLA). The processor 710 may integrate one or a combination of several of the following: Central Processing Unit (CPU), Graphics Processing Unit (GPU), and modem. The CPU primarily handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the displayed content; and the modem handles wireless communication. It is understood that the modem may also not be integrated into the processor and may be implemented using a separate communication chip.

[0120] According to one aspect of the embodiments of this application, a computer program product or computer program is provided, the computer program product or computer program including computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, causing the computer device to perform the methods of any of the above embodiments.

[0121] According to one aspect of this application, a computer-readable storage medium is also provided, which may be included in the electronic device described in the above embodiments; or it may exist independently and not assembled into the electronic device. The computer-readable storage medium carries computer-readable instructions that, when executed by a processor, implement the methods in any of the above embodiments.

[0122] It should be noted that the computer-readable medium shown in the embodiments of this application can be a computer-readable signal medium, a computer-readable storage medium, or any combination of the two. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this application, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such transmitted data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. The computer-readable signal medium can also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to wireless, wired, etc., or any suitable combination thereof.

[0123] The units described in the embodiments of this application can be implemented in software or hardware, and the described units can also be located in a processor. The names of these units do not necessarily limit the specific unit itself.

[0124] It should be noted that although several modules or units for the device used to perform actions have been mentioned in the detailed description above, this division is not mandatory. In fact, according to the embodiments of this application, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0125] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. Each block in a flowchart or block diagram may represent a module, segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0126] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

Claims

1. A method for detecting copied speech, characterized in that, The method includes: Acquire the speech data to be detected, and perform active speech detection on the speech data to be detected; Speech recognition is performed on the active speech in the speech data to be detected to identify at least two speech segments in the speech data to be detected, and the at least two speech segments have the same speech content. The at least two speech segments are subjected to duplicate speech detection to obtain duplicate speech detection results, wherein the duplicate speech detection results include the at least two speech segments as duplicate speech segments; Perform duplicate speech detection on the at least two speech segments to obtain duplicate speech detection results, including: Feature extraction is performed on the at least two speech segments to obtain the feature sequences corresponding to each of the at least two speech segments; Determine the ratio between the feature sequences corresponding to each of the at least two speech segments; If the ratio is a constant value, then based on the constant value, it is determined that the at least two speech segments were obtained by copying and pasting; If the difference between the ratio and the constant value is less than a preset difference, then based on the preset difference, it is determined that the at least two speech segments were obtained by copying, scaling, jittering, and pasting. The constant value includes a third constant value and a fourth constant value, and the preset difference includes a first preset difference and a second preset difference, including: If the difference between the ratio and the third constant value is less than the first preset difference, then it is determined that the first speech segment of the at least two speech segments is obtained by sequentially copying, scaling, and jittering the second speech segment of the at least two speech segments before pasting; or If the difference between the ratio and the fourth constant value is less than the second preset difference, then it is determined that the first speech segment of the at least two speech segments is obtained by sequentially copying, jittering, and scaling the second speech segment of the at least two speech segments and then pasting it.

2. The method according to claim 1, characterized in that, The constant value includes a first constant value and a second constant value. If the ratio is equal to the constant value, then determining that the at least two speech segments were obtained by copying and pasting based on the constant value includes: If the ratio is the first constant value, then it is determined that the first speech segment of the at least two speech segments is obtained by copying and pasting the second speech segment of the at least two speech segments; or If the ratio is the second constant value, then it is determined that the first speech segment of the at least two speech segments is obtained by sequentially copying and scaling the second speech segment of the at least two speech segments and then pasting it.

3. The method according to claim 1, characterized in that, The step of performing duplicate speech detection on the at least two speech segments to obtain duplicate speech detection results includes: Feature extraction is performed on the at least two speech segments to obtain the feature sequences corresponding to each of the at least two speech segments; Determine the difference between the feature sequences corresponding to each of the at least two speech segments; If the difference is a fifth constant value, then it is determined that the first speech segment of the at least two speech segments is obtained by copying and pasting the second speech segment of the at least two speech segments; or If the difference is within the preset difference range, then it is determined that the first speech segment of the at least two speech segments is obtained by sequentially copying and jittering the second speech segment of the at least two speech segments and then pasting it.

4. The method according to any one of claims 1-3, characterized in that, Perform duplicate speech detection on the at least two speech segments to obtain duplicate speech detection results, including: Align the feature sequences corresponding to each of the at least two speech segments to determine the alignment distance between the feature sequences corresponding to each of the at least two speech segments; Based on the alignment distance, a target sequence corresponding to the feature sequence of each of the at least two speech segments is determined, wherein the target sequence includes at least one of an overlap difference sequence and an overlap ratio sequence; If the target sequence meets the alignment condition, then duplicate speech detection is performed on the target sequence to obtain the duplicate speech detection result.

5. The method according to any one of claims 1-3, characterized in that, After performing duplicate speech detection on the at least two speech segments to determine whether the at least two speech segments are duplicate speech segments, the method further includes: Determine the timestamp of the copied speech segment in the speech data to be detected, and determine the speech duration of the copied speech segment based on the timestamp; From the copied audio segments, identify target copied audio segments whose audio duration is greater than a preset duration.

6. A device for detecting copied speech, characterized in that, The device includes: The speech data acquisition module is used to acquire speech data to be detected and to perform active speech detection on the speech data to be detected. The speech recognition module is used to perform speech recognition on the active speech in the speech data to be detected, and to determine at least two speech segments in the speech data to be detected, wherein the at least two speech segments have the same speech content; A duplicate speech detection module is used to perform duplicate speech detection on the at least two speech segments and obtain a duplicate speech detection result, wherein the duplicate speech detection result includes the at least two speech segments as duplicate speech segments; Perform duplicate speech detection on the at least two speech segments to obtain duplicate speech detection results, including: Feature extraction is performed on the at least two speech segments to obtain the feature sequences corresponding to each of the at least two speech segments; Determine the ratio between the feature sequences corresponding to each of the at least two speech segments; If the ratio is a constant value, then it is determined that the at least two speech segments were obtained by copying and pasting based on the constant value, wherein the constant value includes a third constant value and a fourth constant value; If the difference between the ratio and the constant value is less than a preset difference, then based on the preset difference, it is determined that the at least two speech segments were obtained by copying, scaling, jittering, and pasting. The preset difference includes a first preset difference and a second preset difference, including: If the difference between the ratio and the third constant value is less than the first preset difference, then it is determined that the first speech segment of the at least two speech segments is obtained by sequentially copying, scaling, and jittering the second speech segment of the at least two speech segments before pasting; or If the difference between the ratio and the fourth constant value is less than the second preset difference, then it is determined that the first speech segment of the at least two speech segments is obtained by sequentially copying, jittering, and scaling the second speech segment of the at least two speech segments and then pasting it.

7. An electronic device, characterized in that, include: processor; A memory storing computer-readable instructions that, when executed by the processor, implement the method as described in any one of claims 1-5.

8. A computer-readable storage medium, characterized in that, It stores computer-readable instructions that, when executed by a processor, implement the method as described in any one of claims 1-5.

Images