Intelligent interaction system for recognizing voiceprint information through environmental sound

Abstract

The invention discloses an intelligent interaction system for recognizing voiceprint information through environmental sound, and relates to the field of man-machine interaction. Comprising an environment sound pickup module which picks up environment sound in a vehicle by using a microphone array and carries out filtering and denoising processing on the sound; the voiceprint feature extraction module is used for converting the sound signal into a spectrogram by using short-time Fourier transform, distinguishing different passengers through different waveform features, and then extracting MDCC features in the sound signal as sound feature vectors; according to the invention, through picking up the environment sound, then extracting the voice data of the passenger, carrying out feature extraction on the voice data again, obtaining the voice content of the passenger, carrying out emotion and intention analysis on the voice content of the passenger, and then carrying out recommendation and related control, the active degree and the automation degree of man-machine interaction are increased, and the experience of the passenger is improved.

Description

Technical Field

[0001] The present invention relates to the field of vehicle-computer interaction, and in particular to an intelligent interaction system that recognizes voiceprint information through ambient sound. Background Art

[0002] The field of human-computer interaction technology focuses on the design, evaluation, implementation and research of the interaction between humans and onboard computer systems.

[0003] After searching, the Chinese patent publication number CN118860154A discloses a contactless human-computer interaction system, which includes an information acquisition module, a contactless human-computer interaction system initialization module, a contactless input command real-time feedback output display module, a contactless command feedback output processing module, a coordinate correction mapping module, a manipulation action definition and recognition module, a spatial depth pressing multi-layer command virtual machine, a natural manipulation command virtual machine, and a contactless force feedback physical simulation module.

[0004] The above patent has the following deficiencies: although it adopts contactless interaction, it still adopts passive interaction, that is, interaction can only be carried out after the driver and passengers input instructions. It cannot actively perceive the needs of the driver and passengers and interact, which reduces the passenger experience.

[0005] To this end, the present invention proposes an intelligent interactive system for recognizing voiceprint information through ambient sound. Summary of the invention

[0006] The purpose of the present invention is to solve the shortcomings of the prior art and to propose an intelligent interactive system for recognizing voiceprint information through ambient sound.

[0007] In order to achieve the above object, the present invention adopts the following technical solutions: An intelligent interactive system for identifying voiceprint information through ambient sound, comprising: Ambient sound pickup module, which uses a microphone array to pick up ambient sound in the car and performs filtering and denoising on the sound; The voiceprint feature extraction module uses short-time Fourier transform to convert the sound signal into a spectrum, distinguishes different passengers by different waveform features, and then extracts the MDCC features in the sound signal as the sound feature vector; The emotion and intention analysis module analyzes the picked-up sound signals to obtain the emotions and intentions of passengers; Database, including local database and virtual database, the virtual database is connected to big data through the network, used to obtain information from the big data network in real time, and the local database is used to store downloaded information and local voice information; The recommendation module uses a multi-dimensional data fusion algorithm to determine the recommended content based on the emotions and intentions of passengers obtained by the emotion and intention analysis module; The instruction generation and execution module generates control instructions according to the recommended content and controls the corresponding area of ​​the vehicle for execution.

[0008] Preferably: in the voiceprint feature extraction module, the mathematical expression of short-time Fourier transform is: , where x(t) is the sound signal to be analyzed, w(t) is the window function which is used to limit the time range of the analysis, and t is the time variable which represents the center position of the window function. is the angular frequency, which represents the frequency component of the signal, through the sliding window function On the signal x(t), WTFT performs a local Fourier transform on the signal t at each time point to obtain the frequency component amplitude and phase information.

[0009] Preferably: in the voiceprint feature extraction module, according to the frequency components of the sound Then draw the spectrum graph, and then extract the characteristics of different passengers through any one of frequency distribution, amplitude change, spectrum centroid, and spectrum roll-off point.

[0010] Preferably: the voiceprint feature extraction module in the voiceprint feature extraction module is based on the frequency components of the sound Then draw the spectrum diagram, and then extract the characteristics of different passengers through the comprehensive algorithm of frequency distribution, amplitude change, spectrum centroid, and spectrum roll-off point, which specifically includes the following steps: A1: Extract the temporal sound interval of the same passenger in the sound signal through frequency distribution, amplitude change, spectrum centroid, and spectrum roll-off point. ; A2: Set the weights of frequency distribution, amplitude change, spectrum centroid, and spectrum roll-off point, which are ; A3: The last updated feature interval of the passenger is ,in , .

[0011] Preferably: in step A2, the weight is determined by a precision test, and the test includes the following steps: A21: Select any segment of voice conversation with multiple passengers as model input; A22: Determine the standard time sequence sound interval of one of the passengers based on the conversation content ; A23: Use frequency distribution, amplitude change, spectrum centroid, and spectrum roll-off point methods to extract features from speech and obtain the time series interval of the passenger in step A22 , where i represents the frequency distribution or amplitude variation or spectrum centroid or spectrum roll-off point method; A24: Then according to the formula , i represents the frequency distribution or amplitude change or spectrum centroid or spectrum roll-off point method, and the corresponding for Any one of them.

[0012] Preferably: the logic of the emotion and intention analysis module includes the following steps: B1: Obtain the passenger's voice time sequence interval and extract the voice of the corresponding time sequence in the voice signal as voice input; B2: Recognize speech into text signals; B3: Text signal splitting: Split text signals according to sentences and parts of speech. Sentences include subjects, predicates, and objects. Parts of speech include verbs, adjectives, and modal particles. B4: Analyze passengers’ emotions and intentions based on sentence information and parts of speech.

[0013] Preferably: in step B4, analyzing the passenger intention includes the following steps: B41: Multiple emotion and intent data blocks are pre-established, each emotion and intent data corresponds to one or more sentences and part-of-speech combinations, and each sentence and part-of-speech combination can be included in multiple emotion and intent data; B42: Get the passenger's sentences and part-of-speech combinations based on text signal segmentation , Represents the passenger's sentence and the i-th combination of parts of speech, and obtains the frequency of occurrence of each combination ; B43: Then, the matching degree between the passenger and each emotion and intention data is calculated according to the matching degree logic, and the one with the highest matching degree is selected as the passenger's emotional intention.

[0014] Preferably: in the step B43, the calculation formula is: .

[0015] Preferably: in the step B43, S is the matching degree, n is the total number of passenger sentences and part-of-speech combination types extracted, m is the number of combinations of the target emotion and intention data that are identical to the passenger sentences and part-of-speech combination types, For the same combination frequency; Or in the step B43, S is the matching degree, n is the total number of combination types included in the target emotion and intention data, m is the number of the same combination types included in the target emotion and intention data and the passenger's sentence and part-of-speech combination types, The same combination frequency.

[0016] Preferably: the recommendation content of the recommendation module includes light color, brightness control, music type, and volume control.

[0017] Preferably, it is applied to the adjustment of lights and music in coaches, subways and buses.

[0018] The beneficial effects of the present invention are: 1. The present invention picks up the ambient sound, then extracts the passenger's voice data, extracts features from the voice data again, obtains the passenger's voice content, and then analyzes the passenger's voice content for emotions and intentions, and then makes recommendations and related controls, thereby increasing the initiative and automation of human-computer interaction and enhancing the passenger experience.

[0019] 2. The present invention uses multiple features to extract and classify passenger features, and fuses the multiple features in a weight-based manner, thereby increasing the accuracy of extracting and separating sound signals of different passengers in the voice signal.

[0020] 3. The present invention, on the basis of weight-based passenger feature extraction, sets the weight based on accuracy testing. On the one hand, it can combine multiple methods to jointly determine and increase accuracy. On the other hand, based on accuracy control, it can also make the method with higher accuracy in this field have higher decision power, thereby further increasing accuracy.

[0021] 4. The present invention adopts a matching degree calculation method for emotion and intention analysis, and the matching can be calculated based on the passenger extraction result or can be matched based on the emotion and intention analysis data block, so that the matching result and the extraction result have a high degree of fit, further improving the driving experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 This is an architecture diagram of an intelligent interactive system proposed by the present invention that recognizes voiceprint information through ambient sound; Figure 2 This is a feature extraction flow chart of an intelligent interactive system proposed by the present invention that recognizes voiceprint information through ambient sound; Figure 3 This is a logical flow chart of the emotion and intention analysis module of an intelligent interactive system for recognizing voiceprint information through ambient sound proposed by the present invention. DETAILED DESCRIPTION

[0023] The technical solution of the present invention is further described in detail below in conjunction with specific implementation methods.

[0024] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", and "set" should be understood in a broad sense, for example, they can be fixedly connected or set, or detachably connected or set, or integrally connected or set. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0025] Example 1

[0026] An intelligent interactive system for identifying voiceprint information through ambient sound, comprising: Ambient sound pickup module, which uses a microphone array to pick up ambient sound in the car and performs filtering and denoising on the sound; The voiceprint feature extraction module uses short-time Fourier transform to convert the sound signal into a spectrum, distinguishes different passengers by different waveform features, and then extracts the MDCC features in the sound signal as the sound feature vector; The emotion and intention analysis module analyzes the picked-up sound signals to obtain the emotions and intentions of passengers; Database, including local database and virtual database, the virtual database is connected to big data through the network, used to obtain information from the big data network in real time, and the local database is used to store downloaded information and local voice information; The recommendation module uses a multi-dimensional data fusion algorithm to determine the recommended content based on the emotions and intentions of passengers obtained by the emotion and intention analysis module; The instruction generation and execution module generates control instructions according to the recommended content and controls the corresponding area of ​​the vehicle for execution.

[0027] Its workflow is: S1: The ambient sound pickup module uses a microphone array to pick up the ambient sound in the car, and filters and denoises the sound, and transmits the sound signal to the voiceprint feature extraction module after processing; S2: The voiceprint feature extraction module uses short-time Fourier transform to convert the sound signal into a spectrogram, distinguishes different passengers by different waveform features, then extracts the MDCC features in the sound signal as the sound feature vector, and transmits the processed data to the emotion and intention analysis module; S3: The emotion and intention analysis module analyzes the received sound signals, obtains the emotions and intentions of the passengers, and transmits the emotions and intentions to the recommendation module; S4: The recommendation module uses a multi-dimensional data fusion algorithm to determine the recommended content based on the emotions and intentions of the passengers obtained by the emotion and intention analysis module, and transmits the recommended content to the instruction generation and execution module; S5: The instruction generation and execution module generates a control instruction according to the recommended content, and controls the corresponding area of ​​the vehicle to execute it.

[0028] The present invention picks up the ambient sound, then extracts the passenger's voice data, extracts features of the voice data again, obtains the passenger's voice content, and then analyzes the passenger's voice content for emotions and intentions, and then makes recommendations and related controls, thereby increasing the initiative and automation of human-computer interaction and enhancing the driving experience.

[0029] Example 2

[0030] An intelligent interactive system for identifying voiceprint information through ambient sound, comprising: Ambient sound pickup module, which uses a microphone array to pick up ambient sound in the car and performs filtering and denoising on the sound; The voiceprint feature extraction module uses short-time Fourier transform to convert the sound signal into a spectrum, distinguishes different passengers by different waveform features, and then extracts the MDCC features in the sound signal as the sound feature vector; The emotion and intention analysis module analyzes the picked-up sound signals to obtain the emotions and intentions of passengers; Database, including local database and virtual database, the virtual database is connected to big data through the network, used to obtain information from the big data network in real time, and the local database is used to store downloaded information and local voice information; The recommendation module uses a multi-dimensional data fusion algorithm to determine the recommended content based on the emotions and intentions of passengers obtained by the emotion and intention analysis module; The instruction generation and execution module generates control instructions according to the recommended content and controls the corresponding area of ​​the vehicle for execution.

[0031] In the voiceprint feature extraction module, the mathematical expression of short-time Fourier transform is: , where x(t) is the sound signal to be analyzed, w(t) is the window function which is used to limit the time range of the analysis, and t is the time variable which represents the center position of the window function. is the angular frequency, which represents the frequency component of the signal, through the sliding window function On the signal x(t), WTFT performs a local Fourier transform on the signal t at each time point to obtain the frequency component amplitude and phase information.

[0032] In the voiceprint feature extraction module, according to the frequency components of the sound Then draw the spectrum graph, and then extract the characteristics of different passengers through any one of frequency distribution, amplitude change, spectrum centroid, and spectrum roll-off point.

[0033] Example 3

[0034] An intelligent interactive system for identifying voiceprint information through ambient sound, comprising: Ambient sound pickup module, which uses a microphone array to pick up ambient sound in the car and performs filtering and denoising on the sound; The voiceprint feature extraction module uses short-time Fourier transform to convert the sound signal into a spectrum, distinguishes different passengers by different waveform features, and then extracts the MDCC features in the sound signal as the sound feature vector; The emotion and intention analysis module analyzes the picked-up sound signals to obtain the emotions and intentions of passengers; Database, including local database and virtual database, the virtual database is connected to big data through the network, used to obtain information from the big data network in real time, and the local database is used to store downloaded information and local voice information; The recommendation module uses a multi-dimensional data fusion algorithm to determine the recommended content based on the emotions and intentions of passengers obtained by the emotion and intention analysis module; The instruction generation and execution module generates control instructions according to the recommended content and controls the corresponding area of ​​the vehicle for execution.

[0035] In the voiceprint feature extraction module, the mathematical expression of short-time Fourier transform is: , where x(t) is the sound signal to be analyzed, w(t) is the window function which is used to limit the time range of the analysis, and t is the time variable which represents the center position of the window function. is the angular frequency, which represents the frequency component of the signal, through the sliding window function On the signal x(t), WTFT performs a local Fourier transform on the signal t at each time point to obtain the frequency component amplitude and phase information.

[0036] The voiceprint feature extraction module is a module that extracts the voiceprint features according to the frequency components of the sound. Then draw the spectrum diagram, and then extract the characteristics of different passengers through the comprehensive algorithm of frequency distribution, amplitude change, spectrum centroid, and spectrum roll-off point, which specifically includes the following steps: A1: Extract the temporal sound interval of the same passenger in the sound signal through frequency distribution, amplitude change, spectrum centroid, and spectrum roll-off point. ; A2: Set the weights of frequency distribution, amplitude change, spectrum centroid, and spectrum roll-off point, which are ; A3: The last updated feature interval of the passenger is ,in , .

[0037] The present invention uses multiple features to extract and classify passenger features, and fuses the multiple features in a weight-based manner, thereby increasing the accuracy of extracting and separating sound signals of different passengers in the voice signal.

[0038] Example 4

[0039] An intelligent interactive system for identifying voiceprint information through ambient sound, comprising: Ambient sound pickup module, which uses a microphone array to pick up ambient sound in the car and performs filtering and denoising on the sound; The voiceprint feature extraction module uses short-time Fourier transform to convert the sound signal into a spectrum, distinguishes different passengers by different waveform features, and then extracts the MDCC features in the sound signal as the sound feature vector; The emotion and intention analysis module analyzes the picked-up sound signals to obtain the emotions and intentions of passengers; Database, including local database and virtual database, the virtual database is connected to big data through the network, used to obtain information from the big data network in real time, and the local database is used to store downloaded information and local voice information; The recommendation module uses a multi-dimensional data fusion algorithm to determine the recommended content based on the emotions and intentions of passengers obtained by the emotion and intention analysis module; The instruction generation and execution module generates control instructions according to the recommended content and controls the corresponding area of ​​the vehicle for execution.

[0040] In the voiceprint feature extraction module, the mathematical expression of short-time Fourier transform is: , where x(t) is the sound signal to be analyzed, w(t) is the window function which is used to limit the time range of the analysis, and t is the time variable which represents the center position of the window function. is the angular frequency, which represents the frequency component of the signal, through the sliding window function On the signal x(t), WTFT performs a local Fourier transform on the signal t at each time point to obtain the frequency component amplitude and phase information.

[0041] The voiceprint feature extraction module is a module that extracts the voiceprint features according to the frequency components of the sound. After that, the spectrum is drawn, and then the characteristics of different passengers are extracted through the frequency distribution, amplitude change, spectrum centroid, and spectrum roll-off point comprehensive algorithm, which specifically includes the following steps: A1: Extract the temporal sound interval of the same passenger in the sound signal through frequency distribution, amplitude change, spectrum centroid, and spectrum roll-off point. ; A2: Set the weights of frequency distribution, amplitude change, spectrum centroid, and spectrum roll-off point, which are ; A3: The last updated feature interval of the passenger is ,in , .

[0042] In step A2, the weight is determined by a precision test, which includes the following steps: A21: Select any segment of voice conversation with multiple passengers as model input; A22: Determine the standard time sequence sound interval of one of the passengers based on the conversation content ; A23: Use frequency distribution, amplitude change, spectrum centroid, and spectrum roll-off point methods to extract features from speech and obtain the time series interval of the passenger in step A22 , where i represents the frequency distribution or amplitude variation or spectrum centroid or spectrum roll-off point method; A24: Then according to the formula , i represents the frequency distribution or amplitude change or spectrum centroid or spectrum roll-off point method, and the corresponding for Any one of them.

[0043] The present invention, on the basis of weight-based passenger feature extraction, sets the weight based on accuracy test. On the one hand, it can combine multiple methods to jointly determine and increase the accuracy. On the other hand, based on accuracy control, it can also make the method with higher accuracy in this field have higher decision power, thereby further increasing the accuracy.

[0044] Example 5

[0045] An intelligent interactive system for identifying voiceprint information through ambient sound, comprising: Ambient sound pickup module, which uses a microphone array to pick up ambient sound in the car and performs filtering and denoising on the sound; The voiceprint feature extraction module uses short-time Fourier transform to convert the sound signal into a spectrum, distinguishes different passengers by different waveform features, and then extracts the MDCC features in the sound signal as the sound feature vector; The emotion and intention analysis module analyzes the picked-up sound signals to obtain the emotions and intentions of passengers; Database, including local database and virtual database, the virtual database is connected to big data through the network, used to obtain information from the big data network in real time, and the local database is used to store downloaded information and local voice information; The recommendation module uses a multi-dimensional data fusion algorithm to determine the recommended content based on the emotions and intentions of passengers obtained by the emotion and intention analysis module; The instruction generation and execution module generates control instructions according to the recommended content and controls the corresponding area of ​​the vehicle for execution.

[0046] In the voiceprint feature extraction module, the mathematical expression of short-time Fourier transform is: , where x(t) is the sound signal to be analyzed, w(t) is the window function which is used to limit the time range of the analysis, and t is the time variable which represents the center position of the window function. is the angular frequency, which represents the frequency component of the signal, through the sliding window function On the signal x(t), WTFT performs a local Fourier transform on the signal t at each time point to obtain the frequency component amplitude and phase information.

[0047] The voiceprint feature extraction module is a module that extracts the voiceprint features according to the frequency components of the sound. After that, the spectrum is drawn, and then the characteristics of different passengers are extracted through the frequency distribution, amplitude change, spectrum centroid, and spectrum roll-off point comprehensive algorithm, which specifically includes the following steps: A1: Extract the temporal sound interval of the same passenger in the sound signal through frequency distribution, amplitude change, spectrum centroid, and spectrum roll-off point. ; A2: Set the weights of frequency distribution, amplitude change, spectrum centroid, and spectrum roll-off point, which are ; A3: The last updated feature interval of the passenger is ,in , .

[0048] In step A2, the weight is determined by a precision test, which includes the following steps: A21: Select any segment of voice conversation with multiple passengers as model input; A22: Determine the standard time sequence sound interval of one of the passengers based on the conversation content ; A23: Use frequency distribution, amplitude change, spectrum centroid, and spectrum roll-off point methods to extract features from speech and obtain the time series interval of the passenger in step A22 , where i represents the frequency distribution or amplitude variation or spectrum centroid or spectrum roll-off point method; A24: Then according to the formula , i represents the frequency distribution or amplitude change or spectrum centroid or spectrum roll-off point method, and the corresponding for Any one of them.

[0049] The logic of the emotion and intention analysis module includes the following steps: B1: Obtain the passenger's voice time sequence interval and extract the voice of the corresponding time sequence in the voice signal as voice input; B2: Recognize speech into text signals; B3: Text signal splitting: Split text signals according to sentences and parts of speech. Sentences include subjects, predicates, and objects. Parts of speech include verbs, adjectives, and modal particles. B4: Analyze passengers’ emotions and intentions based on sentence information and parts of speech.

[0050] In step B4, analyzing the passenger intention includes the following steps: B41: Multiple emotion and intent data blocks are pre-established, each emotion and intent data corresponds to one or more sentences and part-of-speech combinations, and each sentence and part-of-speech combination can be included in multiple emotion and intent data; B42: Get the passenger's sentences and part-of-speech combinations based on text signal segmentation , Represents the passenger's sentence and the i-th combination of parts of speech, and obtains the frequency of occurrence of each combination ; B43: Then, the matching degree between the passenger and each emotion and intention data is calculated according to the matching degree logic, and the one with the highest matching degree is selected as the passenger's emotional intention.

[0051] In step B43, the calculation formula is: .

[0052] In the step B43, S is the matching degree, n is the total number of passenger sentences and part-of-speech combination types extracted, m is the number of combinations of the target emotion and intention data that are identical to the passenger sentences and part-of-speech combination types, The same combination frequency.

[0053] Or in the step B43, S is the matching degree, n is the total number of combination types included in the target emotion and intention data, m is the number of the same combination types included in the target emotion and intention data and the passenger's sentence and part-of-speech combination types, The same combination frequency.

[0054] The recommendation content of the recommendation module includes light color, brightness control, music type, and volume control.

[0055] The present invention adopts a matching degree calculation method for emotion and intention analysis, and the matching can be calculated based on the passenger extraction result or can be matched based on the emotion and intention analysis data block, so that the matching result and the extraction result have a high degree of fit, further improving the passenger experience.

[0056] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. An intelligent interactive system for identifying voiceprint information through ambient sound, characterized in that: include: Ambient sound pickup module, which uses a microphone array to pick up ambient sound in the car and performs filtering and denoising on the sound; The voiceprint feature extraction module uses short-time Fourier transform to convert the sound signal into a spectrum, distinguishes different passengers by different waveform features, and then extracts the MDCC features in the sound signal as the sound feature vector; The emotion and intention analysis module analyzes the picked-up sound signals to obtain the emotions and intentions of passengers; Database, including local database and virtual database, the virtual database is connected to big data through the network, used to obtain information from the big data network in real time, and the local database is used to store downloaded information and local voice information; The recommendation module uses a multi-dimensional data fusion algorithm to determine the recommended content based on the emotions and intentions of passengers obtained by the emotion and intention analysis module; The command generation and execution module generates control commands based on the recommended content and controls the corresponding area of ​​the vehicle for execution; The recommendation module recommends content including light color, brightness control, music type, and volume control; Its workflow is: S1: The ambient sound pickup module uses a microphone array to pick up the ambient sound in the car, and filters and denoises the sound, and transmits the sound signal to the voiceprint feature extraction module after processing; S2: The voiceprint feature extraction module uses short-time Fourier transform to convert the sound signal into a spectrogram, distinguishes different passengers by different waveform features, then extracts the MDCC features in the sound signal as the sound feature vector, and transmits the processed data to the emotion and intention analysis module; S3: The emotion and intention analysis module analyzes the received sound signals, obtains the emotions and intentions of the passengers, and transmits the emotions and intentions to the recommendation module; S4: The recommendation module uses a multi-dimensional data fusion algorithm to determine the recommended content based on the emotions and intentions of the passengers obtained by the emotion and intention analysis module, and transmits the recommended content to the instruction generation and execution module; S5: The instruction generation and execution module generates a control instruction according to the recommended content, and controls the corresponding area of ​​the vehicle to execute it.

2. The intelligent interactive system for identifying voiceprint information through ambient sound according to claim 1, characterized in that: In the voiceprint feature extraction module, the mathematical expression of short-time Fourier transform is: , where x(t) is the sound signal to be analyzed, w(t) is the window function which is used to limit the time range of the analysis, and t is the time variable which represents the center position of the window function. is the angular frequency, which represents the frequency component of the signal, through the sliding window function On the signal x(t), WTFT performs a local Fourier transform on the signal t at each time point to obtain the frequency component amplitude and phase information.

3. The intelligent interactive system for identifying voiceprint information through ambient sound according to claim 1, characterized in that: In the voiceprint feature extraction module, according to the frequency components of the sound Then draw the spectrum graph, and then extract the characteristics of different passengers through any one of frequency distribution, amplitude change, spectrum centroid, and spectrum roll-off point.

4. The intelligent interactive system for identifying voiceprint information through ambient sound according to claim 1, characterized in that: The voiceprint feature extraction module is a module that extracts the voiceprint features according to the frequency components of the sound. After that, the spectrum is drawn, and then the characteristics of different passengers are extracted through the frequency distribution, amplitude change, spectrum centroid, and spectrum roll-off point comprehensive algorithm, which specifically includes the following steps: A1: Extract the temporal sound interval of the same passenger in the sound signal through frequency distribution, amplitude change, spectrum centroid, and spectrum roll-off point. ; A2: Set the weights of frequency distribution, amplitude change, spectrum centroid, and spectrum roll-off point, which are ; A3: The last updated feature interval of the passenger is ,in , .

5. The intelligent interactive system for identifying voiceprint information through ambient sound according to claim 4, characterized in that: In step A2, the weight is determined by a precision test, which includes the following steps: A21: Select any segment of voice conversation with multiple passengers as model input; A22: Determine the standard time sequence sound interval of one of the passengers based on the conversation content ; A23: Use frequency distribution, amplitude change, spectrum centroid, and spectrum roll-off point methods to extract features from speech and obtain the time series interval of the passenger in step A22 , where i represents the frequency distribution or amplitude variation or spectrum centroid or spectrum roll-off point method; A24: Then according to the formula , i represents the frequency distribution or amplitude change or spectrum centroid or spectrum roll-off point method, and the corresponding for Any one of them.

6. The intelligent interactive system for identifying voiceprint information through ambient sound according to claim 1, characterized in that: The logic of the emotion and intention analysis module includes the following steps: B1: Obtain the passenger's voice time sequence interval and extract the voice of the corresponding time sequence in the voice signal as voice input; B2: Recognize speech into text signals; B3: Text signal splitting: Split text signals according to sentences and parts of speech. Sentences include subjects, predicates, and objects. Parts of speech include verbs, adjectives, and modal particles. B4: Analyze passengers’ emotions and intentions based on sentence information and parts of speech.

7. The intelligent interactive system for identifying voiceprint information through ambient sound according to claim 6, characterized in that: In step B4, analyzing the passenger intention includes the following steps: B41: Multiple emotion and intent data blocks are pre-established, each emotion and intent data corresponds to one or more sentences and part-of-speech combinations, and each sentence and part-of-speech combination can be included in multiple emotion and intent data; B42: Get the passenger's sentences and part-of-speech combinations based on text signal segmentation , Represents the passenger's sentence and the i-th combination of parts of speech, and obtains the frequency of occurrence of each combination ; B43: Then, the matching degree between the passenger and each emotion and intention data is calculated according to the matching logic, and the one with the highest matching degree is selected as the passenger's emotional intention.

8. The intelligent interactive system for identifying voiceprint information through ambient sound according to claim 7, characterized in that: In step B43, the calculation formula is: .

9. The intelligent interactive system for identifying voiceprint information through ambient sound according to claim 8, characterized in that: In the step B43, S is the matching degree, n is the total number of passenger sentences and part-of-speech combination types extracted, m is the number of combinations of the target emotion and intention data that are identical to the passenger sentences and part-of-speech combination types, For the same combination frequency; Or in the step B43, S is the matching degree, n is the total number of combination types included in the target emotion and intention data, m is the number of the same combination types included in the target emotion and intention data and the passenger's sentence and part-of-speech combination types, The same combination frequency.

10. An intelligent interactive system for recognizing voiceprint information through ambient sound according to any one of claims 1 to 9, characterized in that: It is used for adjusting lights and music sounds in coaches, subways and buses.

Images