Vehicle-mounted voice dialogue method and system and vehicle machine

Abstract

The invention provides a vehicle-mounted voice dialogue method and system and a vehicle machine. The method comprises the following steps that: a central processing unit judges whether user voice input received by a microphone arranged on a vehicle comprises an awakening word or not, and awakens a vehicle-mounted voice dialogue process if the user voice input comprises the awakening word; the audio digital signal processing chip performs echo cancellation and noise reduction to generate input audio subjected to echo cancellation and noise reduction; performing voice recognition on the input audio by a graphics processor to generate a voice recognition text; executing semantic understanding by the central processing unit; the central processing unit determines the intention of the user according to the current semantics and the dialogue historical information, and obtains reply information aiming at the intention of the user; a speech synthesis is performed by the embedded neural network processor according to the reply information, a reply audio is generated, and the reply audio is transmitted to a speaker disposed within the vehicle. The operation load of the central processing unit is reduced, and better voice interaction experience is provided under the limited hardware condition.

Description

Technical Field

[0001] The embodiments of the present invention generally relate to the technical field of in-vehicle information systems, and more specifically, to an in-vehicle voice dialogue method, system and vehicle computer. Background Art

[0002] The hardware performance of the in-vehicle infotainment system host (hereinafter referred to as the car computer) is limited. In order to provide users with rich multi-scenario and intelligent services, there are often a large number of service software competing for the hardware resources of the car computer.

[0003] The general solution for voice engines in in-car entertainment systems is to rely entirely on the computing power of the central processing unit (CPU). As the functions of in-car entertainment systems become richer and more complex, the CPU load is overwhelmed. During the development process, in some cases, the CPU load exceeds 90% for a long time, resulting in slow response and lag in the voice dialogue system.

[0004] Therefore, the hardware resources allocated by the car computer to the intelligent voice dialogue system often cannot meet the needs of the voice engine, affecting the performance of the voice system under actual load. It is necessary to find an optimization solution to solve the problem of insufficient system resources. Summary of the invention

[0005] In order to solve the above-mentioned problems in the prior art, in the first aspect, an embodiment of the present invention provides an in-vehicle voice conversation method, the method comprising: a central processing unit determines whether a user voice input received by a microphone arranged on a vehicle includes a wake-up word, and wakes up the in-vehicle voice conversation process when the user voice input includes the wake-up word; an audio digital signal processing chip performs echo cancellation and noise reduction on the user voice input received via the microphone to generate input audio after echo cancellation and noise reduction, and transmits the input audio after echo cancellation and noise reduction to a graphics processor; the graphics processor performs speech recognition on the input audio after echo cancellation and noise reduction to generate The invention relates to a method for generating a speech recognition text and transmitting the speech recognition text to the central processing unit; the central processing unit performs semantic understanding on the speech recognition text and identifies the current semantics of the speech recognition text of the current conversation; the central processing unit determines the user intention according to the current semantics and conversation history information, calls the application corresponding to the user intention to obtain reply information for the user intention, and transmits the reply information to the embedded neural network processor; the embedded neural network processor performs speech synthesis according to the reply information, generates reply audio, and transmits the reply audio to a speaker arranged in the vehicle so that the speaker plays the reply audio.

[0006] In some embodiments, the method also includes: receiving a wake-up indication signal by the central processor, wherein the wake-up indication signal comes from a voice wake-up hard key or a voice wake-up soft key on the vehicle computer; in response to receiving the wake-up indication signal, waking up the in-vehicle voice conversation process by the central processor.

[0007] In some embodiments, the audio digital signal processing chip performs echo cancellation and noise reduction on the user voice input received via the microphone, including: the audio digital signal processing chip simulates the channel environment of the echo in the user voice input by adjusting the parameters of the adaptive filter to generate an echo signal estimation; the audio digital signal processing chip subtracts the echo signal estimation from the signal of the user voice input to generate an echo-cancelled signal.

[0008] In some embodiments, the audio digital signal processing chip performs echo cancellation and noise reduction on the user voice input received via the microphone, including: the audio digital signal processing chip determines the direction and amplitude of the external noise source from the user voice input; the audio digital signal processing chip controls the mute speaker to emit a sound wave with the opposite direction and the same amplitude as the external noise source.

[0009] In the second aspect, an embodiment of the present invention proposes an in-vehicle voice dialogue system, the system comprising a central processing unit, an audio digital signal processing chip, a graphics processor and an embedded neural network processor. The central processing unit is used to: determine whether the user voice input received by the microphone set on the vehicle includes a wake-up word, and if the user voice input includes the wake-up word, wake up the in-vehicle voice dialogue process; semantically understand the voice recognition text transmitted by the graphics processor, and identify the current semantics of the voice recognition text of the current dialogue; determine the user intention according to the current semantics and dialogue history information, call the application corresponding to the user intention to obtain reply information for the user intention, and transmit the reply information to the embedded neural network processor. The audio digital signal processing chip is used to: perform echo cancellation and noise reduction on the user voice input received via the microphone, generate input audio after echo cancellation and noise reduction, and transmit the input audio after echo cancellation and noise reduction to the graphics processor. The graphics processor is used to: perform voice recognition on the input audio after echo cancellation and noise reduction, generate voice recognition text, and transmit the voice recognition text to the central processing unit. The embedded neural network processor is used to perform speech synthesis according to the reply information, generate reply audio, and transmit the reply audio to a speaker arranged in the vehicle so that the speaker plays the reply audio.

[0010] In some embodiments, the central processor is also used to: receive a wake-up indication signal, wherein the wake-up indication signal comes from a voice wake-up hard key or a voice wake-up soft key on the vehicle computer; and wake up the vehicle voice dialogue process in response to receiving the wake-up indication signal.

[0011] In some embodiments, the audio digital signal processing chip is further used to: simulate the channel environment of the echo in the user voice input by adjusting the parameters of the adaptive filter to generate an echo signal estimation; subtract the echo signal estimation from the signal of the user voice input to generate a signal after echo cancellation.

[0012] In some embodiments, the audio digital signal processing chip is further used to: determine the direction and amplitude of the external noise source from the user voice input; and control the mute speaker to emit a sound wave with the opposite direction and the same amplitude as the external noise source.

[0013] In some embodiments, the central processing unit, the audio digital signal processing chip, the graphics processor, and the embedded neural network processor are integrated on the same system-level chip.

[0014] In a third aspect, an embodiment of the present invention provides a vehicle computer, including the vehicle-mounted voice dialogue system described in any of the above embodiments.

[0015] The existing voice engine solution is that all modules of the voice engine run calculations in the CPU, all relying on the performance and computing power resources of the CPU. The above heterogeneous solution proposed in the embodiment of the present invention splits the voice engine into several parts, among which the wake-up, dialogue management, and semantic understanding modules use CPU calculations, echo cancellation / noise reduction is processed by Audio DSP chip, speech recognition uses GPU calculations, and speech synthesis (TTS) uses NPU calculations, forming a CPU-GPU-NPU-DSP heterogeneous solution for the vehicle-mounted voice engine.

[0016] GPU is currently mainly used for navigation map rendering, and its usage rate in the development of in-vehicle entertainment systems is relatively low, so the load is usually low. The current mainstream automotive-grade system chips all have GPU processors with strong processing capabilities. Compared with CPU, GPU has the characteristics of more cores and fewer control units, so it is more suitable for computationally intensive and data-parallel programs. The deep learning algorithm model meets the above two characteristics, and the ASR module is implemented using the deep learning model.

[0017] Text-to-speech (TTS) is implemented based on a deep learning algorithm based on a neural network model, and is therefore very suitable for NPU chip processing.

[0018] In order to improve the quality of Bluetooth calls, in-vehicle entertainment systems often integrate Audio DSP chips with professional echo cancellation and noise reduction functions. The voice conversation function can reuse this capability, thereby reducing the load on other hardware.

[0019] Splitting each module instead of using only the CPU and GPU also further achieves the goal of balancing the load of each chip as much as possible.

[0020] The above heterogeneous solution proposed in the embodiment of the present invention fully utilizes the capabilities of other processing units of the host SoC chip, and solves the bottleneck problem that the performance of the voice system is limited by system resources. Compared with the existing use of the CPU for all in-vehicle voice dialogue processing, the embodiment of the present invention reduces the CPU computing load and reduces the amount of computing that the CPU needs to perform to within the range that the CPU can bear. Therefore, it can better cope with the complex in-vehicle entertainment system resource environment and provide a better voice interaction experience under limited hardware conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The above and other objects, features and advantages of the embodiments of the present invention will become readily understood by reading the following detailed description with reference to the accompanying drawings. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, in which:

[0022] Figure 1 A flow chart of a vehicle-mounted voice dialogue method according to an embodiment of the present invention is shown;

[0023] Figure 2 A schematic diagram of an in-vehicle voice dialogue system according to an embodiment of the present invention is shown.

[0024] In the drawings, the same or corresponding reference numerals represent the same or corresponding parts. DETAILED DESCRIPTION

[0025] The principle and spirit of the present invention will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are provided only to enable those skilled in the art to better understand and implement the present invention, and are not intended to limit the scope of the present invention in any way.

[0026] In one aspect, an embodiment of the present invention provides a method for in-vehicle voice dialogue. Figure 1 , which shows a flow chart of an in-vehicle voice dialogue method 100 according to an embodiment of the present invention. The method 100 includes S101-S106.

[0027] In step S101, the central processing unit (CPU) determines whether the user voice input received by the microphone installed in the vehicle includes a wake-up word, such as "Hello, XX (name of the vehicle system)", and wakes up the vehicle voice dialogue process if the user voice input includes the wake-up word. Generally, the CPU is the core processor responsible for calculation and control of the computer system.

[0028] As an embodiment of the present invention, the wake-up process may also be: the central processing unit receives a wake-up indication signal, wherein the wake-up indication signal comes from a voice wake-up hard key or a voice wake-up soft key on the vehicle computer; in response to receiving the wake-up indication signal, the central processing unit wakes up the in-vehicle voice dialogue process.

[0029] In step S102, an audio (Audio) digital signal processing (Digital Signal Processor, DSP) chip performs echo cancellation and noise reduction (ECNR) on the user voice input received via the microphone, generates input audio after echo cancellation and noise reduction, and transmits the input audio after echo cancellation and noise reduction to the graphics processor.

[0030] Echo exists when the microphone picks up the signal from the speaker in duplex mode, such as when the car's audio system recognizes the lyrics when the car's head unit plays music.

[0031] As an embodiment of the present invention, echo cancellation can be performed in the following manner: the audio digital signal processing chip simulates the channel environment of the echo in the user voice input by adjusting the parameters of the adaptive filter to generate an echo signal estimation; the audio digital signal processing chip subtracts the echo signal estimation from the signal of the user voice input to generate an echo-cancelled signal.

[0032] Echo cancellation is usually implemented using an adaptive filter, that is, a filter with adjustable parameters is designed, and the filter parameters are adjusted through an adaptive algorithm to simulate the channel environment where the echo is generated, thereby estimating the size of the echo signal, and then subtracting this estimated value from the received signal to cancel the echo. For example, if the car's audio system is playing a song, the song's sound wave data should be eliminated from the recorded voice command audio data to prevent the played song from affecting the recognition of the correct voice command.

[0033] As an embodiment of the present invention, noise reduction can be performed in the following manner: the audio digital signal processing chip determines the direction and amplitude of the external noise source from the user's voice input; the audio digital signal processing chip controls the mute speaker to emit a sound wave with the same amplitude and opposite direction to the external noise source. Through the noise reduction system, that is, through the mute speaker to emit a sound wave with the same amplitude and opposite direction to the external noise source, the two sound waves are superimposed, thereby reducing the noise. Example: The in-vehicle voice system generally needs to reduce the impact of tire noise, engine noise, wind noise, etc. on the accuracy of voice recognition.

[0034] In step S103, the graphics processing unit (GPU) performs speech recognition (ASR) on the input audio that has been echo-cancelled and denoised, generates speech recognition text, and transmits the speech recognition text to the central processing unit. Generally, the GPU is a processor that is mainly responsible for image and graphics related operations, and can also perform general computing. Speech recognition is a technology that converts human speech into text. The machine converts the voice signal into corresponding text or command through the process of recognition and understanding. For example: the user says a command such as "help me", and the speech recognition correctly converts the audio command into the corresponding text.

[0035] In step S104, the CPU performs Natural Language Understanding (NLU) on the speech recognition text to identify the current semantics of the speech recognition text of the current conversation. Semantic understanding is used to simulate the human language communication process, so that the computer can understand and use the natural language of human society such as Chinese, English, etc., to achieve natural language communication between man and machine. In the in-vehicle voice system, NLU can classify the ASR result text of the user's voice input into various sub-fields, such as checking the weather, booking air tickets, etc., and identify the user's specific intentions. For example: for the instruction "help me book a flight to Shanghai tomorrow", semantic understanding will identify the functional domain as booking a flight, the time as tomorrow, and the destination as Shanghai.

[0036] In step S105, the CPU determines the user intention based on the current semantics and dialogue history information, that is, performs dialogue management (DM), and then calls the application corresponding to the user intention to obtain reply information for the user intention, and transmits the reply information to the embedded neural network processor.

[0037] The dialogue management decides how to respond to the user at this moment based on the dialogue history information. Currently, the most common application is task-driven multi-round dialogue, where users come with clear purposes such as ordering food or tickets. User needs are relatively complex and have many restrictions, which may require multiple rounds of statements. On the one hand, users can constantly modify or improve their needs during the dialogue process. On the other hand, when the user's stated needs are not specific or clear enough, the machine can also help users find satisfactory results by asking, clarifying or confirming. For example: After the user inquires about Beijing's weather today by asking "How is Beijing's weather?", he asks again "What about tomorrow?" The dialogue management system can automatically complete the user's intention to inquire about Beijing's weather tomorrow based on the context.

[0038] In step S106, an embedded neural network processor (NPU) performs text-to-speech synthesis (TTS) based on the reply information, generates reply audio, and transmits the reply audio to a speaker installed in the vehicle so that the speaker plays the reply audio. The NPU is a processor specially designed to accelerate the operation of the neural network model.

[0039] The purpose of speech generation is to transform text information into natural and fluent human language audio. The processing process is mainly divided into three steps: ① Disassemble the text to obtain the duration and frequency changes of phonemes. ② The combination of characters forms words, forms pauses in a natural way, and confirms the pronunciation of polyphones in the form of phrases. ③ Combine the speaker's speaking habits, pronunciation characteristics, accent characteristics, etc. to form a voice with obvious human characteristics. Example: The voice dialogue system has generated a reply text to feedback the user's instructions. The speech generation module converts the text into natural and fluent simulated audio for playback.

[0040] The existing voice engine solution is that all modules of the voice engine run calculations in the CPU, all relying on the performance and computing power resources of the CPU. The above heterogeneous solution proposed in the embodiment of the present invention splits the voice engine into several parts, among which the wake-up, dialogue management, and semantic understanding modules use CPU calculations, echo cancellation / noise reduction is processed by Audio DSP chip, speech recognition uses GPU calculations, and speech synthesis (TTS) uses NPU calculations, forming a CPU-GPU-NPU-DSP heterogeneous solution for the vehicle-mounted voice engine.

[0041] GPU is currently mainly used for navigation map rendering, and its usage rate in the development of in-vehicle entertainment systems is relatively low, so the load is usually low. The current mainstream automotive-grade system chips all have GPU processors with strong processing capabilities. Compared with CPU, GPU has the characteristics of more cores and fewer control units, so it is more suitable for computationally intensive and data-parallel programs. The deep learning algorithm model meets the above two characteristics, and the ASR module is implemented using the deep learning model.

[0042] Text-to-speech (TTS) is implemented based on a deep learning algorithm based on a neural network model, and is therefore very suitable for NPU chip processing.

[0043] In order to improve the quality of Bluetooth calls, in-vehicle entertainment systems often integrate Audio DSP chips with professional echo cancellation and noise reduction functions. The voice conversation function can reuse this capability, thereby reducing the load on other hardware.

[0044] Splitting each module instead of using only the CPU and GPU also further achieves the goal of balancing the load of each chip as much as possible.

[0045] See Table 1 below, which shows the results of the optimized computing power test of a certain project of the in-vehicle voice system. The data in the table represents the CPU computing power, and the numerical unit is DMIPS. Taking Qualcomm Snapdragon 8155 chip as an example, the total CPU computing power is about 95k DMIPS.

[0046] Table 1. Example of CPU computing power test results before and after optimization

[0047]

[0048] It can be seen that the above heterogeneous solution proposed in the embodiment of the present invention gives full play to the capabilities of other processing units of the host SoC chip, and solves the bottleneck problem that the performance of the voice system is limited by system resources. Compared with the existing use of the CPU for all in-vehicle voice dialogue processing, the embodiment of the present invention reduces the CPU computing load and reduces the amount of calculations that the CPU needs to perform to within the range that the CPU can bear. Therefore, it can better cope with the complex in-vehicle entertainment system resource environment and provide a better voice interaction experience under limited hardware conditions.

[0049] On the other hand, an embodiment of the present invention provides a vehicle-mounted voice dialogue system. Figure 2 , which shows a schematic diagram of a vehicle-mounted voice dialogue system according to an embodiment of the present invention. The system includes a central processing unit, an audio digital signal processing chip, a graphics processor and an embedded neural network processor.

[0050] The central processing unit is used to: determine whether the user voice input received by the microphone installed on the vehicle includes a wake-up word, and if the user voice input includes the wake-up word, wake up the vehicle voice conversation process; perform semantic understanding on the voice recognition text transmitted by the graphics processor, and identify the current semantics of the voice recognition text of the current conversation; determine the user intention based on the current semantics and conversation history information, call the application corresponding to the user intention to obtain reply information for the user intention, and transmit the reply information to the embedded neural network processor.

[0051] The audio digital signal processing chip is used to perform echo cancellation and noise reduction on the user voice input received via the microphone, generate input audio after echo cancellation and noise reduction, and transmit the input audio after echo cancellation and noise reduction to the graphics processor.

[0052] The graphics processor is used for: performing speech recognition on the input audio after echo cancellation and noise reduction, generating speech recognition text, and transmitting the speech recognition text to the central processor.

[0053] The embedded neural network processor is used to perform speech synthesis according to the reply information, generate reply audio, and transmit the reply audio to a speaker arranged in the vehicle so that the speaker plays the reply audio.

[0054] As an embodiment of the present invention, the central processor can also be used to: receive a wake-up indication signal, where the wake-up indication signal comes from a voice wake-up hard key or a voice wake-up soft key on the vehicle computer; in response to receiving the wake-up indication signal, wake up the vehicle voice dialogue process.

[0055] As an embodiment of the present invention, the audio digital signal processing chip can be further used to: simulate the channel environment of the echo in the user voice input by adjusting the parameters of the adaptive filter to generate an echo signal estimation; subtract the echo signal estimation from the signal of the user voice input to generate a signal after echo cancellation.

[0056] As an embodiment of the present invention, the audio digital signal processing chip can be further used to: determine the direction and amplitude of the external noise source from the user voice input; control the mute speaker to emit sound waves with the opposite direction and the same amplitude as the external noise source.

[0057] As an embodiment of the present invention, a central processing unit, an audio digital signal processing chip, a graphics processor and an embedded neural network processor are integrated on the same system on chip (SoC).

[0058] In yet another aspect, an embodiment of the present invention provides a vehicle computer, comprising the in-vehicle voice dialogue system described according to any of the above embodiments.

[0059] The in-vehicle voice dialogue method, system and vehicle computer proposed in the embodiments of the present invention use heterogeneous design to split the voice engine that originally relied entirely on CPU computing power into several parts, give full play to the capabilities of other processing units of the host SoC chip, such as GPU, NPU, and DSP, to solve the bottleneck problem that the performance of the voice system is limited by system resources, and provide a better voice interaction experience under limited hardware conditions.

[0060] For the purpose of illustration, the foregoing description of the embodiments of the present invention has been given, which is not exhaustive nor intended to limit the present invention to the disclosed exact form. It will be appreciated by those skilled in the art that various changes can be made without departing from the scope of the present invention, and that the elements therein can be replaced with equivalents. In addition, without departing from the basic scope of the present invention, many modifications can be made so that specific situations or materials are adapted to the teachings of the present invention. Therefore, the present invention is not intended to be limited to the specific embodiments disclosed as the best mode for realizing the present invention, and the present invention will include all embodiments falling within the scope of the appended claims.

Claims

1. A vehicle-mounted voice dialogue method, characterized in that: The method comprises: The central processor determines whether the user voice input received by the microphone disposed on the vehicle includes a wake-up word, and wakes up the vehicle voice dialogue process if the user voice input includes the wake-up word; The audio digital signal processing chip performs echo cancellation and noise reduction on the user voice input received via the microphone, generates input audio after echo cancellation and noise reduction, and transmits the input audio after echo cancellation and noise reduction to the graphics processor; The graphics processor performs speech recognition on the echo-cancelled and denoised input audio to generate speech recognition text, and transmits the speech recognition text to the central processor; The central processor performs semantic understanding on the speech recognition text to identify the current semantics of the speech recognition text of the current conversation; The central processor determines the user intention according to the current semantics and the conversation history information, calls the application corresponding to the user intention to obtain the reply information for the user intention, and transmits the reply information to the embedded neural network processor; The embedded neural network processor performs speech synthesis according to the reply information to generate reply audio, and transmits the reply audio to a speaker arranged in the vehicle so that the speaker plays the reply audio.

2. The method according to claim 1, characterized in that The method further comprises: The central processor receives a wake-up indication signal, wherein the wake-up indication signal comes from a voice wake-up hard key or a voice wake-up soft key on the vehicle computer; In response to receiving the wake-up indication signal, the central processor wakes up the in-vehicle voice dialogue process.

3. The method according to claim 1, characterized in that The audio digital signal processing chip performs echo cancellation and noise reduction on the user voice input received via the microphone, including: The audio digital signal processing chip simulates the channel environment of the echo in the user's voice input by adjusting the parameters of the adaptive filter to generate an echo signal estimation; The audio digital signal processing chip subtracts the echo signal estimate from the signal of the user voice input to generate a signal after echo cancellation.

4. The method according to claim 1, characterized in that: The audio digital signal processing chip performs echo cancellation and noise reduction on the user voice input received via the microphone, including: The audio digital signal processing chip determines the direction and amplitude of the external noise source from the user voice input; The audio digital signal processing chip controls the mute speaker to emit a sound wave in a direction opposite to that of the external noise source and with the same amplitude.

5. A vehicle-mounted voice dialogue system, characterized in that: The system includes a central processing unit, an audio digital signal processing chip, a graphics processor and an embedded neural network processor, wherein: The central processor is used to: determine whether the user voice input received by the microphone arranged on the vehicle includes a wake-up word, and if the user voice input includes the wake-up word, wake up the vehicle voice dialogue process; perform semantic understanding on the voice recognition text transmitted by the graphics processor, and identify the current semantics of the voice recognition text of the current dialogue; determine the user intention according to the current semantics and dialogue history information, call the application corresponding to the user intention to obtain the reply information for the user intention, and transmit the reply information to the embedded neural network processor; The audio digital signal processing chip is used to: perform echo cancellation and noise reduction on the user voice input received via the microphone, generate input audio after echo cancellation and noise reduction, and transmit the input audio after echo cancellation and noise reduction to the graphics processor; The graphics processor is used to: perform speech recognition on the echo-cancelled and denoised input audio, generate speech recognition text, and transmit the speech recognition text to the central processor; The embedded neural network processor is used to perform speech synthesis according to the reply information, generate reply audio, and transmit the reply audio to a speaker arranged in the vehicle so that the speaker plays the reply audio.

6. The system according to claim 5, characterized in that The central processing unit is also used for: Receiving a wake-up indication signal, wherein the wake-up indication signal comes from a voice wake-up hard key or a voice wake-up soft key on the vehicle computer; In response to receiving the wake-up indication signal, the in-vehicle voice dialogue process is woken up.

7. The system according to claim 5, characterized in that The audio digital signal processing chip is further used for: By adjusting the parameters of the adaptive filter, the channel environment of the echo in the user's voice input is simulated to generate an echo signal estimation; The echo signal estimate is subtracted from the signal of the user voice input to generate an echo-cancelled signal.

8. The system according to claim 5, characterized in that The audio digital signal processing chip is further used for: Determining the direction and magnitude of an external noise source from the user voice input; Control the noise canceling speaker to emit sound waves in the opposite direction and with the same amplitude as the external noise source.

9. The system according to claim 5, characterized in that The central processing unit, the audio digital signal processing chip, the graphics processor and the embedded neural network processor are integrated on the same system-level chip.

10. A vehicle computer, characterized in that: Comprising an in-vehicle voice dialogue system according to any one of claims 5-9.

Images