Intelligent device wake-up method and apparatus

Abstract

The application discloses an intelligent device wake-up method and device in the field of artificial intelligence, and the method comprises the following steps: receiving a first voice containing a wake-up word, the wake-up word being a word set in advance for waking up an intelligent device; acquiring the confidence of the wake-up word contained in the first voice; determining a first scene and a wake-up threshold corresponding to the first scene according to an interactive intention of a previous round corresponding to the first voice, wherein the interactive intention of the previous round is an interactive intention corresponding to a second voice, the second voice being a voice received before the first voice, the second voice containing an instructive sentence for controlling the intelligent device to perform an operation, and the first scene being a scene corresponding to the interactive intention of the previous round; and in the case that the confidence is greater than the wake-up threshold, waking up the intelligent device. The above technical scheme realizes the determination of a corresponding scene and a wake-up threshold corresponding to the scene based on the interactive intention of the previous round, so as to determine whether to wake up the intelligent device according to the confidence and the wake-up threshold.

Description

Technical Field

[0001] This invention relates to human-computer interaction technology in the field of artificial intelligence, and particularly to a method and apparatus for waking up a smart device. Background Technology

[0002] Human-computer interaction, as a research direction in the field of artificial intelligence, has evolved from interactions via keyboards, mice, and touchscreens, represented by personal computers (PCs) and smartphones, to voice interaction, represented by intelligent dialogue systems such as mobile voice assistants, smart speakers, smart screens, and smart vehicles. Before voice interaction, users need to wake up the smart device using voice commands, such as "Hi, Xiaoyi," to bring it from sleep mode to working mode so that it can process user commands properly. Therefore, the quality of voice wake-up significantly impacts the user experience of voice interaction.

[0003] Smart devices are pre-configured with wake-up models. The system inputs the user's voice into the wake-up model to obtain the confidence level of the wake-up words included in the user's voice. If the confidence level of the wake-up word is greater than the wake-up threshold, the smart device is awakened and enters a state where it can process user commands normally. Commonly used wake-up models include statistical models such as Hidden Markov Models (HMMs) and Deep Neural Networks (DNNs). Currently, most solutions use a fixed wake-up threshold. If the wake-up threshold is set too high, in crowded scenarios such as user gatherings, inconsistent voice tones and mixed sound sources from various sources can easily make it difficult to wake up the smart device. If the wake-up threshold is set too low, in quieter scenarios such as when users are resting, false wake-ups are likely. Therefore, a fixed wake-up threshold is insufficient to meet the needs of different users in different environments, resulting in a poor user experience. Summary of the Invention

[0004] This invention proposes a method and apparatus for waking up smart devices, addressing the problem that a fixed wake-up threshold is insufficient to meet the wake-up threshold requirements of different users in different environments, resulting in a poor user experience. The technical solution is as follows:

[0005] In a first aspect, embodiments of this application propose a method for waking up a smart device, applied to a smart device, the method comprising:

[0006] Receive a first voice message containing a wake word, wherein the wake word is a pre-set word used to wake up the smart device;

[0007] Obtain the confidence level of the wake word contained in the first speech;

[0008] Based on the interaction intent of the previous round corresponding to the first voice, determine the first scenario and the wake-up threshold corresponding to the first scenario; wherein, the interaction intent of the previous round is the interaction intent corresponding to the second voice, the second voice is the voice received before receiving the first voice, the second voice contains the instruction statement for controlling the smart device to perform operations, and the first scenario is the scenario corresponding to the interaction intent of the previous round.

[0009] The smart device is woken up when the confidence level is greater than the wake-up threshold.

[0010] The above technical solution enables the determination of the corresponding scenario and wake-up threshold based on the interaction intent of the previous round, thereby determining whether to wake up the smart device based on the confidence level and the wake-up threshold.

[0011] In one possible implementation, the wake-up threshold includes a threshold that is automatically updated based on historical interaction behavior data, wherein the historical interaction behavior data includes the number of “scene triggers” and the number of “scene-wake-up event pairs”, where the number of “scene triggers” is the cumulative number of triggers in a certain scene, and the number of “scene-wake-up event pairs” is the cumulative number of wake-ups of the smart device in a certain scene.

[0012] In one possible implementation, the method further includes:

[0013] When the smart device is woken up, update the number of "scene triggers" and the number of "scene-wake-up event pairs" corresponding to the first scene;

[0014] Based on the number of "scene triggers" and "scene-wake-up event pairs" corresponding to the updated first scene, the wake-up threshold corresponding to the first scene is updated, thereby achieving the effect of automatically adjusting the wake-up threshold based on historical interaction behavior data.

[0015] In one possible implementation, the wake-up threshold corresponding to the first scenario is updated based on the updated number of "scenario triggers" and the number of "scenario-wake-up event pairs," specifically including:

[0016] Update the wake-up probability corresponding to the first scenario based on the number of "scenario triggers" and the number of "scenario-wake-up event pairs" corresponding to the first scenario after the update;

[0017] The wake-up threshold corresponding to the first scenario is updated based on the wake-up probability of the smart device corresponding to the updated first scenario.

[0018] In this possible implementation, the wake-up probability of the first scenario is automatically adjusted based on historical interaction behavior data, and then the wake-up threshold of the first scenario is automatically adjusted based on the adjusted wake-up probability.

[0019] In one possible implementation, the wake-up probability corresponding to the first scenario described above is calculated based on the following formula:

[0020]

[0021] Wherein, P(w|c1) represents the wake-up probability corresponding to the first scenario, n(c1,w) represents the number of "scenario-wake-up event pairs" corresponding to the first scenario, and n(c1) represents the number of "scenario triggers" corresponding to the first scenario.

[0022] In one possible implementation, the wake-up threshold corresponding to the first scenario described above is calculated based on the following formula:

[0023]

[0024] Among them, the The wake-up threshold used to represent the first scenario, The values ​​are used to represent the wake-up threshold corresponding to the daily standby scenario, where the wake-up threshold for the daily standby scenario is 0.5. P(w|c1) represents the wake-up probability corresponding to the first scenario, P(w|c0) represents the wake-up probability corresponding to the daily standby scenario, and α represents the adjustment range.

[0025] In one possible implementation, the method further includes:

[0026] Record the first wake-up event corresponding to the first voice, and determine the first moment of the first wake-up event corresponding to the first voice.

[0027] In one possible implementation, the method further includes:

[0028] Determine the time difference between the first moment and the second moment, where the second moment is the moment when the interactive intent corresponding to the second voice is determined;

[0029] If the time difference is less than a preset time threshold, the first voice and the second voice are determined to be in the same session.

[0030] Secondly, embodiments of the present invention also provide a smart device wake-up device, comprising at least one processor, the processor being configured to execute instructions stored in a memory to cause the aforementioned communication device to perform:

[0031] The method described in the first aspect and the steps in various possible implementations.

[0032] Thirdly, embodiments of the present invention provide a computer program product containing instructions that, when run on a computer, cause the computer to execute:

[0033] The method described in the first aspect and the steps in various possible implementations.

[0034] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the steps of the method described in the first aspect and various possible implementations are performed. Attached Figure Description

[0035] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0036] Figure 1 A schematic diagram of an artificial intelligence main framework provided in an embodiment of the present invention;

[0037] Figure 2 This is a system architecture diagram for automatically adjusting wake-up sensitivity provided in an embodiment of the present invention;

[0038] Figure 3 A detailed structural diagram of the sensitivity control module provided in an embodiment of the present invention;

[0039] Figure 4 A flowchart illustrating a smart device wake-up method provided in an embodiment of the present invention;

[0040] Figure 5 This is another flowchart illustrating a smart device wake-up method provided in an embodiment of the present invention. Detailed Implementation

[0041] The technical solutions of the embodiments of the present invention will now be described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0042] It should be noted that the term "and / or" in this application is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. The terms "first" and "second," etc., in the specification and claims of this invention are used to distinguish different objects, not to describe a specific order of objects. For example, "first speech" and "second speech," etc., are used to distinguish different speech, not to describe a specific order of target objects. In this invention, the words "exemplary," "for example," or "e.g.," are used to indicate examples, illustrations, or explanations. Any embodiment or design described as "exemplary," "for example," or "e.g.," in this invention should not be construed as superior to other embodiments or designs. Specifically, the use of "exemplary" or "e.g.," etc., is intended to present related concepts in a specific manner. In the description of this invention, unless otherwise stated, "multiple" means two or more.

[0043] Figure 1 A schematic diagram of an artificial intelligence framework is shown, which describes the overall workflow of an artificial intelligence system and is applicable to general artificial intelligence domain needs.

[0044] The above-mentioned artificial intelligence framework will be elaborated from two dimensions: "intelligent information chain" (horizontal axis) and "IT value chain" (vertical axis).

[0045] The "intelligent information chain" reflects a series of processes from data acquisition to processing. For example, it could be a general process of intelligent information perception, intelligent information representation and formation, intelligent reasoning, intelligent decision-making, and intelligent execution and output. In this process, data undergoes a condensation process of "data—information—knowledge—wisdom."

[0046] The "IT value chain" reflects the value that artificial intelligence brings to the information technology industry, from the underlying infrastructure of artificial intelligence, information (provided and processed by technology) to the industrial ecosystem of systems.

[0047] (1) Infrastructure:

[0048] Infrastructure provides computing power to support artificial intelligence systems, enabling communication with the external world and providing support through a basic platform. This communication occurs through sensors; computing power is provided by intelligent chips (hardware acceleration chips such as CPUs, NPUs, GPUs, ASICs, and FPGAs); and the basic platform includes distributed computing frameworks and related platform guarantees and support, which may include cloud storage and computing, interconnected networks, etc. For example, sensors communicate with the outside world to acquire data, and this data is provided to intelligent chips in the distributed computing system provided by the basic platform for computation.

[0049] (2) Data

[0050] The data at the next layer of infrastructure is used to represent the data sources in the field of artificial intelligence. The data involves graphics, images, voice, text, and IoT data from traditional devices, including business data from existing systems and sensor data such as force, displacement, liquid level, temperature, and humidity.

[0051] (3) Data processing

[0052] Data processing typically includes methods such as data training, machine learning, deep learning, search, reasoning, and decision-making.

[0053] Among them, machine learning and deep learning can perform intelligent information modeling, extraction, preprocessing, and training on data, including symbolization and formalization.

[0054] Reasoning refers to the process in which, in a computer or intelligent system, the machine thinks and solves problems by simulating human intelligent reasoning, based on reasoning control strategies and using formalized information. Typical functions include search and matching.

[0055] Decision-making refers to the process of making decisions based on intelligent information after reasoning, and it typically provides functions such as classification, sorting, and prediction.

[0056] (4) General ability

[0057] After the data processing mentioned above, the results of the data processing can be used to form some general capabilities, such as algorithms or a general system, for example, translation, text analysis, computer vision processing, speech recognition, image recognition, etc.

[0058] (5) Smart Products and Industry Applications

[0059] Intelligent products and industry applications refer to products and applications of artificial intelligence systems in various fields. They encapsulate overall artificial intelligence solutions, productize intelligent information decision-making, and realize practical applications. Their application areas mainly include: intelligent manufacturing, intelligent transportation, smart home, intelligent healthcare, intelligent security, autonomous driving, safe city, and intelligent terminals.

[0060] To facilitate the explanation of this solution, the definitions of terms that may be involved in the embodiments of this invention are given first:

[0061] (1) Wake-up word: A preset word used to wake up a smart device. The smart device can respond to a user's voice containing the wake-up word, activating it from sleep mode. Wake-up words can be, for example, "Xiaoyi," "Hi," etc. "wait.

[0062] (2) Wake-up Sensitivity: The degree to which a smart device responds to voice input from a user containing a wake-up word. In this embodiment of the invention, a higher wake-up sensitivity means that the smart device is easier to wake up, and a lower wake-up sensitivity means that the smart device is less likely to be woken up. Wake-up sensitivity can be in the form of, for example, "high," "medium," "low," or a quantified numerical value. It should be understood that those skilled in the art can also use logic opposite to that in the embodiments of the present invention to set the wake-up sensitivity, without exceeding the scope of this application.

[0063] (3) Wake-up Threshold: This characterizes the trigger condition for a smart device to be woken up by a wake-up word. Different wake-up words may have the same or different wake-up thresholds. The wake-up threshold can be measured numerically. That is, the wake-up threshold is the minimum threshold value for a smart device to be woken up by a wake-up word. Generally, when the confidence level of the wake-up word contained in the user's voice received by the smart device is greater than the above wake-up threshold, the smart device is woken up; when the confidence level of the wake-up word contained in the user's voice received by the smart device is less than the above wake-up threshold, the smart device is not woken up. It should be understood that those skilled in the art can also use logic opposite to that of the embodiments of the present invention to set the wake-up threshold, and all such methods will not exceed the scope of this application.

[0064] (4) Wake-up Event: This refers to the event in which a smart device is woken up when the confidence level of the wake-up word contained in the user's voice is greater than the wake-up threshold corresponding to the scene mapped to the previous round of interaction intent. For example, if the confidence level of the wake-up word contained in the user's voice is 0.6, which is greater than the wake-up threshold of 0.4 corresponding to the scene mapped to the previous round of interaction intent, then the smart device is woken up. A wake-up event can be represented as, for example, "Previous round of interaction intent - play music, wake up", where "play music" is the previous round of interaction intent.

[0065] (5) Session: A series of user voice inputs that are relatively continuous in time. In this embodiment of the invention, a preset time threshold can be used to determine whether two consecutive user voice inputs are in the same session. For example, if the time difference between the wake-up event corresponding to the later user voice input received by the sensitivity control module 207 and the interaction intent corresponding to the earlier user voice input received is less than the preset time threshold, then the later user voice input and the earlier user voice input can be considered to be in the same session.

[0066] (6) Interaction Intent: Represents the intention that the user wants to express through voice. For example, the user's voice "play a popular song randomly" maps to the interaction intent "play music". Another example is the user's voice "play a children's story", which maps to the interaction intent "play children's content".

[0067] (7) Scene: This refers to the conversational environment of the smart device as determined by the user's voice input. The user's voice input can be used to determine the corresponding interaction intent. According to the intent-scene mapping table (i.e., Table 1), the interaction intent can be mapped to the corresponding scene, such as "Entertainment", "Children" and "Daily Standby" as shown in Table 1.

[0068] (8) Number of “Scene Triggers”: Represents the number of times a certain scene is triggered. The number of “scene triggers” can be accumulated and updated. For example, assuming that the original value of the number of “scene triggers” n(c0) for the “daily standby” scene is 299, after the “daily standby” scene is triggered again, the number of “scene triggers” n(c0) is updated to 299 + 1 = 300.

[0069] (9) Number of “Scene-Wake-up Event Pairs”: Represents the number of times the smart device is woken up when a certain scene is triggered. The number of “Scene-Wake-up Event Pairs” can be accumulated, recorded, and updated. For example, the original value of the number of “scene triggers” n(c0) for the “Daily Standby” scene is 299. Assuming that these 299 “Daily Standby” scenes include 99 scenes where the smart device is woken up and 200 scenes where the smart device is not woken up, then the original value of the number of times the smart device is woken up in the “Daily Standby” scene is n(c0,w) = 99. If the smart device is woken up when the “Daily Standby” scene is triggered again, then the number of times the smart device is woken up in the “Daily Standby” scene is updated to n(c0,w) = 99 + 1 = 100.

[0070] (10) Interaction behavior data: including the number of "scene-wake-up event pairs", the number of "scene triggers", and the interaction information table. The number of "scene-wake-up event pairs" and the number of "scene triggers" can be used to calculate the wake-up probability of a smart device in a certain scene, and then update the wake-up threshold corresponding to the certain scene. The interaction information table is used to record the time when the sensitivity control module 207 receives a wake-up event and whether the received event is a wake-up event, as well as the interaction intent received by the sensitivity control module 207, the scene mapped to the interaction intent, and the time when the interaction intent is received.

[0071] Unless otherwise defined, all technical and scientific terms used in this application have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. In case of any inconsistency, the meaning as set forth in the specification or derived from the content described herein shall prevail. Furthermore, the terminology used in this application is for the purpose of describing embodiments of the invention only and is not intended to limit the scope of this application.

[0072] Next, we will first introduce the possible application scenarios and technical problems solved by the solutions provided in the embodiments of the present invention.

[0073] The widespread application of artificial intelligence systems across various fields has led to the increasing prevalence of human-computer interaction. For example, in the field of smart homes, users can control home appliances via voice recognition. Before voice interaction, users need to wake up the smart device using voice commands such as "Hi, Xiaoyi," to bring it from sleep mode to working mode, allowing it to process user commands correctly. Therefore, the effectiveness of voice wake-up significantly impacts the user experience of voice interaction. If the wake-up threshold is set too high (i.e., the wake-up sensitivity is set too low), in crowded scenarios, inconsistent voice tones and mixed audio sources from various sources can easily make it difficult to wake up the smart device. Conversely, if the wake-up threshold is set too low (i.e., the wake-up sensitivity is set too high), false wake-ups are likely to occur in quiet environments.

[0074] In one possible implementation, the smart device can dynamically adjust its wake-up threshold based on preset parameter information from its surrounding environment, thus meeting the user's wake-up threshold requirements in different usage scenarios. However, this method does not consider user behavior; the preset parameter information from the surrounding environment is the same regardless of the user's voice input, whether it contains a wake-up word like "I am Xiaoyi" or a command to control the smart device like "Play a popular song." This preset parameter information can be user activity data. Optionally, this user activity data can be, for example, network traffic data detected by the router the smart device is connected to. As the above analysis shows, this solution only adjusts the wake-up threshold based on preset parameter information from the surrounding environment, and cannot adjust it according to different user behaviors, thus impairing some user experiences.

[0075] In another possible implementation, the smart device can adjust the wake-up threshold of at least one wake-up word by acquiring its wake-up threshold and counting the number of abnormal wake-ups corresponding to that word. This makes the wake-up threshold more adaptable to changes in the terminal's environment. Abnormal wake-ups refer to a discrepancy between the actual wake-up status and the expected wake-up status of the smart device. For example, this could be a false wake-up or a non-wake-up. A false wake-up occurs when the smart device is actually awakened, but its expected wake-up status is that it should not be awakened. A non-wake-up occurs when the smart device is not actually awakened, but its expected wake-up status is that it should be awakened. This method does not consider user behavior and context, and may lead to fluctuations between false wake-ups and non-wake-ups.

[0076] To address this, this invention proposes a method for waking up a smart device. This method automatically identifies the scenario based on the user's previous interaction intent within the same session; automatically calculates the probability of waking up the smart device in that scenario based on historical interaction data; and automatically adjusts the wake-up threshold based on the wake-up probability of the smart device in that scenario, thereby automatically adjusting the wake-up sensitivity. When waking up a smart device, the above technical solution can automatically identify information such as the previous interaction intent corresponding to the user's voice containing the wake word and the scenario mapped to that previous interaction intent, thereby flexibly and automatically adjusting the wake-up threshold. This makes the wake-up threshold more reasonable and better adaptable to changes in the scenario in which the smart device is located, resulting in better performance of the smart device in responding to the wake word and improving the user experience. The above method can be used as follows: Figure 2 The system architecture diagram shown illustrates the implementation of automatically adjusting the wake-up threshold.

[0077] The following is a detailed description of the system architecture diagram for automatically adjusting the wake-up threshold.

[0078] See Figure 2 The above system architecture diagram includes, but is not limited to, a microphone 201, an audio signal acquisition and processing module 202, an audio stream ring buffer 203, a voice assistant application (APP) 204, a voice assistant cloud 205, a wake-up engine 206, and a sensitivity control module 207. A detailed structural diagram of the sensitivity control module 207 is shown below. Figure 3 As shown, it includes, but is not limited to, the interaction behavior statistics submodule 2071 and the scene recognition submodule 2072.

[0079] In the above system architecture, the process is as follows: First, microphone 201 collects the user's voice input. Second, audio signal acquisition and processing module 202 collects the user's voice input from microphone 201, converts it into an audio stream, and sends it to audio stream Ring Buffer 203. Then, wake-up engine 206 obtains the wake-up threshold corresponding to the scene mapped to the previous round of interaction intent from sensitivity control module 207, and wake-up engine 206 monitors the audio stream in audio stream Ring Buffer 203 in real time, and segments it into several voice segments. Wake-up engine 206 performs streaming processing on several voice segments. For example, when wake-up engine 206 processes the first voice segment, if the confidence level of the wake-up word included in the user's voice input is less than the aforementioned wake-up threshold, then it continues to process the second voice segment. At this time, if the confidence level of the wake-up word included in the user's voice input is greater than the aforementioned wake-up threshold, wake-up engine 206 stops processing subsequent voice segments. At this time, wake-up engine 206 wakes up the smart device and sends a wake-up event to sensitivity control module 207. After the smart device is successfully woken up, wake-up engine 206 activates voice assistant APP 204. Next, the voice assistant APP 204 receives voice input from the user via the audio stream Ring Buffer 203. One possible implementation is that the voice assistant APP 204 sends the user's voice input to the voice assistant cloud 205 to request the corresponding interaction intent. It should be noted that the voice assistant cloud 205 can perform Natural Language Understanding (NLU) to obtain the interaction intent corresponding to the user's voice input. Furthermore, after receiving the request from the voice assistant APP 204, the voice assistant cloud 205 returns the interaction intent corresponding to the user's voice input. Finally, the voice assistant APP 204 transmits the interaction intent, such as "play music" or "play children's content," to the sensitivity control module 207. At this point, the sensitivity control module 207 completes the reception of one interaction intent and records information such as the time, scenario, and whether it is a wake-up event. For ease of description, the time when the sensitivity control module 207 receives this interaction intent is referred to as the third moment, and the user's voice input corresponding to this interaction intent is referred to as the first user voice.

[0080] After the sensitivity control module 207 receives an interaction intent, if it receives another wake-up event from the wake-up engine 206, it will determine whether the user's voice input corresponding to the wake-up event is in the same session as the user's voice input corresponding to the previously received interaction intent (the first user's voice). The specific process is as follows:

[0081] When the sensitivity control module 207 receives the next wake-up event from the wake-up engine 206 (denoted as the second user's voice input corresponding to this wake-up event), this moment is designated as the fourth moment. The sensitivity control module 207 compares the fourth moment with the third moment (i.e., the moment when the sensitivity control module 207 last received the interaction intent from the voice assistant APP 204). If the time difference between the fourth moment and the third moment is within a preset time threshold, it is considered that the second user's voice and the first user's voice are in the same session. For example, if the sensitivity control module 207 receives the wake-up event from the wake-up engine 206 again at 13:01:00 on January 24, 2021, and receives the interaction intent from the voice assistant APP 204 at 13:00:00 on January 24, 2021, and the preset time threshold is 3 minutes, since the time difference between the two moments is 1 minute, which is less than the preset time threshold (3 minutes), the sensitivity control module 207 determines that the second user's voice and the first user's voice are in the same session. The above example is only one illustration.

[0082] After the above steps, the interaction behavior statistics submodule 2071 of the sensitivity control module 207 collects user interaction behavior data, which can calculate the wake-up probability of the smart device in each scenario, and thus calculate the wake-up threshold corresponding to each scenario.

[0083] In embodiments of the present invention, such as Figure 2 The system architecture diagram shown implements a method for waking up a smart device. The following will combine... Figure 2 The system architecture diagram shown describes in more detail how to automatically identify scenarios based on the user's previous interaction intent in the same session, automatically calculate the wake-up probability of smart devices in that scenario based on historical interaction behavior data, and automatically adjust the wake-up threshold based on the wake-up probability of smart devices in that scenario.

[0084] like Figure 2 The microphone 201 shown collects the user's third voice, which here refers to the user's voice containing the wake word. For example... Figure 2 The audio signal acquisition and processing module 202 shown converts it into an audio stream and sends it to the audio stream Ring Buffer 203. The wake-up engine 206, from... Figure 2 The sensitivity control module 207 shown queries the wake-up threshold of the current scene. For example, if the sensitivity control module 207 finds that the interaction intent of the previous round in the same session was "standby state," according to the intent-scene mapping table shown in Table 1, it can be determined that the interaction intent "standby state" in the second column maps to the scene "normal standby" in the first column. Therefore, the wake-up threshold for this scene in the third column is 0.5, and the wake-up sensitivity requirement for this scene is "unchanged." Figure 2The wake-up engine 206 shown monitors the audio stream in the Ring Buffer 203 in real time and divides it into several speech segments. The wake-up engine 206 performs streaming processing on the speech segments. When the confidence of the wake word included in the audio stream of a certain speech segment is greater than the wake-up threshold of 0.5 corresponding to the "normal standby" scenario, the wake-up engine 206 stops processing the subsequent speech segments and wakes up the smart device.

[0085] Next, taking the wake word pre-set to "Xiaoyi" as an example, and the user's third voice prompt being "Yiyi, how's the weather today? Xiaoyi, what's the temperature?", we will describe in detail the process by which the wake-up engine 206 determines whether to wake up the smart device based on the user's third voice prompt:

[0086] After the audio signal acquisition and processing module 202 converts the user's third voice into an audio stream, one possible implementation is that the wake-up engine 206 segments the audio stream into audio segments of a preset duration at a preset interval. For example, the preset interval is 1 second, and the preset duration is 5 seconds, meaning 0-5 seconds is the first audio segment, 1-6 seconds is the second audio segment, 2-7 seconds is the third audio segment, and so on, until the entire audio stream is segmented. In this embodiment of the invention, the segmented audio segments can be "Yi Yi" (audio segment A), "Yi, today's weather" (audio segment B), "How's the weather?" (audio segment C), "How are you, Xiao Yi?" (audio segment D), "Xiao Yi, what's the temperature?" (audio segment E), and "What's the temperature?" (audio segment F). In this case, one possible implementation is that the wake-up engine 206 performs streaming processing on audio segments A, B, C, D, E, and F. Specifically, first, the wake-up engine 206 processes voice segment A, outputting a confidence level of 0.2 for the wake-up word "Xiaoyi" contained in the third voice segment, which is less than the aforementioned wake-up threshold of 0.5. Second, the wake-up engine 206 processes voice segment B, outputting a confidence level of 0.3 for the wake-up word "Xiaoyi" contained in the third voice segment, which is less than the aforementioned wake-up threshold of 0.5. Then, the wake-up engine 206 processes voice segment C, outputting a confidence level of 0.2 for the wake-up word "Xiaoyi" contained in the third voice segment, which is less than the aforementioned wake-up threshold of 0.5. Next, the wake-up engine 206 processes voice segment D, outputting a confidence level of 0.6 for the wake-up word "Xiaoyi" contained in the third voice segment, which is greater than the aforementioned wake-up threshold. At this point, the wake-up engine stops processing subsequent voice segments and wakes up the smart device. The aforementioned smart device includes, but is not limited to, smart speakers, smart screens, smart car devices, mobile phones, tablets, laptops, desktop computers, watches, and smart bands.

[0087] Table 1: Intent-Scene Mapping Table

[0088]

[0089] It should be noted that the wake-up threshold for the "normal standby" scenario is preset and remains unchanged, always a fixed value (as shown in Table 1, 0.5 is used as an example in this embodiment). The wake-up thresholds for other scenarios are dynamically changing. The wake-up threshold for a specific scenario is compared with the wake-up threshold for the "normal standby" scenario. If the wake-up threshold for a specific scenario is greater than that for the "normal standby" scenario, it indicates that the wake-up sensitivity needs to be reduced in that scenario; if the wake-up threshold for a specific scenario is less than that for the "normal standby" scenario, it indicates that the wake-up sensitivity needs to be increased in that scenario. For example, as shown in Table 1, the wake-up threshold of 0.6872 for the "Do Not Disturb" scenario is greater than the wake-up threshold of 0.5 for the "normal standby" scenario, indicating that the wake-up sensitivity needs to be reduced in the "Do Not Disturb" scenario; the wake-up threshold of 0.3837 for the "Entertainment" scenario is less than the wake-up threshold of 0.5 for the "normal standby" scenario, indicating that the wake-up sensitivity needs to be increased in the "Entertainment" scenario.

[0090] It should also be noted that multiple interaction intents from the previous round of the same session can be mapped to a scene. As shown in Table 1, the interaction intents "play music", "play audio", and "play game" are mapped to the same scene "entertainment".

[0091] After the smart device is woken up, the wake-up engine 206 is activated. Figure 2 The voice assistant APP 204 shown sends a wake-up event (triggered by a third voice including a wake word) to the sensitivity control module 207 for recording. For example, in one possible implementation, the interaction behavior statistics submodule 2071 of the sensitivity control module 207 (such as...) Figure 3 As shown, the received wake-up event and the time of receipt are written into the interaction information table (e.g., Tables 2 and 3). That is, when a new wake-up event occurs, the interaction information table is updated, and a new wake-up event record is added. The wake-up event record can include information such as time, interaction intent, scenario, and whether it is a wake-up event. For example, Table 3 exemplarily shows an interaction information table obtained after the original interaction information table (Table 2) is updated after a new wake-up event occurs. As shown in Tables 2 and 3, Table 3 adds a wake-up event that occurred on February 20, 2021 at 11:00:01, compared to Table 2. The interaction intent corresponding to this wake-up event is "none," the scenario is "none," and the wake-up event is "yes." This wake-up event can be triggered by user voice that includes a wake word but does not include instructions to control the smart device to perform operations, such as "Hi, Xiaoyi," "Xiaoyi Xiaoyi," or "Xiaoyi, are you there?"

[0092] Next, taking the wake-up event as "previous round of interaction intent - standby state, wake up", and the scenario mapped to the previous round of interaction intent "standby state" as "daily standby" as an example, the process of calculating the wake-up probability of smart devices in the "daily standby" scenario will be described in detail:

[0093] The wake-up engine 206 sends the wake-up event "Previous interaction intent - standby state, wake up" to the sensitivity control module 207. Here, "wake up" in the wake-up event "Previous interaction intent - standby state, wake up" indicates that this event is a wake-up event, and "Previous interaction intent - standby state" means that before this wake-up event, the last interaction intent received by the sensitivity control module 207 from the voice assistant APP 204 was "standby state". The scene recognition submodule 2072 of the sensitivity control module (e.g., Figure 3 As shown, the interaction intent “standby state” is mapped to the scenario “daily standby” according to the intent-scenario mapping table (i.e., Table 1), and the number of “scenario-wake-up event pairs” n(c0,w) under the scenario “daily standby” is updated.

[0094] Assume the original value of n(c0,w) is 500, meaning that before this wake-up event, the smart device was woken up 500 times within a preset time threshold in the "normal standby" scenario. After this wake-up event, the number of times the smart device was woken up within the preset time threshold in the "normal standby" scenario increased by one. Therefore, n(c0,w) is updated to n(c0,w) = 500 + 1 = 501.

[0095] Assume that the original value of the number of "scene triggers" n(c0) in the "normal standby" scenario is 1000, that is, the "normal standby" scenario has occurred 1000 times before this wake-up event; after this wake-up event, the number of times the "normal standby" scenario has occurred has increased by one. Therefore, n(c0) is updated to n(c0) = 1000 + 1 = 1001.

[0096] Thus, the wake-up probability P(w|c) of a smart device in a certain scenario can be obtained. i The calculation formula (1) is as follows:

[0097]

[0098] Where, at this time, c i The scene represents i, which is a positive integer greater than or equal to 0, and the wake-up event represents w.

[0099] When i = 0, the probability of waking up a smart device in the "normal standby" scenario is:

[0100]

[0101] Where c0 represents the scenario of "normal standby" and w represents the wake-up event.

[0102] It should be noted that n(c0)-n(c0,w)=1001-501=500 means that when the "daily standby" scenario is triggered 1001 times, the cumulative number of times the smart device is not woken up within the preset time threshold for each "daily standby" scenario is 500.

[0103] Table 2: Interaction Information Table (1)

[0104] time Interaction Intent Scene Wake-up event 2020.12.01 12:30:05 none none yes 2020.12.01 12:30:25 Play children's content child no 2021.01.30 12:54:08 none none yes 2021.01.30 12:54:23 Play children's content child no

[0105] Table 3: Interaction Information Table (2)

[0106] time Interaction Intent Scene Wake-up event 2020.12.01 12:30:05 none none yes 2020.12.01 12:30:25 Play children's content child no 2021.01.30 12:54:08 none none yes 2021.01.30 12:54:23 Play children's content child no 2021.02.20 11:00:01 none none yes

[0107] Figure 4 This is a flowchart illustrating a smart device wake-up method according to an embodiment of the present invention, which includes steps S402-S420. The flowchart will be described in detail below.

[0108] S402, receives the second voice signal.

[0109] After the voice assistant app 204 is activated, the microphone 201 collects the user's second voice. Here, the second voice refers to the user's voice controlling the smart device to perform operation commands. For example, the second voice could be "Play song A". The audio signal acquisition and processing module 202 collects the user's second voice from the microphone 201, converts it into an audio stream, and sends it to the audio stream Ring Buffer 203. The voice assistant app 204 then sends the audio stream from the audio stream Ring Buffer 203 to... Figure 2 The voice assistant cloud 205 shown obtains the corresponding interaction intent and instructions. These instructions are used to instruct the smart device to perform the operation corresponding to the second voice command.

[0110] S404, Based on the second voice, determine the interaction intent for this round corresponding to the second voice and the first scenario corresponding to the interaction intent for this round.

[0111] Taking the user's second voice command as "play a song for the party" as an example, the voice assistant cloud 205 returns to the voice assistant APP 204 the interaction intent corresponding to the second voice command in this round as "play music", the corresponding command as "play a song", and an address of a song suitable for playing at the party. The address can be, for example, a Uniform Resource Locator (URL).

[0112] The voice assistant app 204 executes the corresponding command, retrieves the song for the party from the URL and plays it, and sends the interaction intent "play music" to the sensitivity control module 207. The interaction behavior statistics submodule 2071 of the sensitivity control module 207 records this interaction intent. That is, when a new interaction intent is received, the interaction information table is updated, and a new record is added to the interaction information table. This record includes the time when the interaction intent was received, the interaction intent, the scene corresponding to the interaction intent, and whether the event corresponding to the interaction intent is a wake-up event. In addition, after receiving the interaction intent from the interaction behavior statistics submodule 2071, the scene recognition submodule 2072 of the sensitivity control module 207 can map the interaction intent to the corresponding scene according to the intent-scene mapping table (i.e., Table 1), and also record the scene corresponding to the interaction intent in the interaction information table.

[0113] Taking the second voice prompt as "play songs for the party" as an example. Assuming the interaction information table before the smart device receives the second voice prompt is Table 3, after the smart device receives the second voice prompt and the sensitivity control module 207 receives the corresponding interaction intent "play music," the interaction information table (i.e., Table 3) will be updated. The updated Table 3 can be, for example, as shown in Table 4. Compared to Table 3, Table 4 has an additional record of the "play music" interaction intent that occurred on February 20, 2021, at 11:00:15. For example, the voice assistant APP 204 sends the current interaction intent "play music" to the sensitivity control module 207. The scene recognition submodule 2072 of the sensitivity control module 207 maps the current interaction intent "play music" to the first scene "entertainment" according to the intent-scene mapping table (i.e., Table 1). Furthermore, the interaction behavior statistics submodule 2071 of the sensitivity control module 207 records in the interaction information table the event that occurred on February 20, 2021 at 11:00:15, with the interaction intent being "play music" and the first scenario being "entertainment". The event referred to here is the event corresponding to user voice commands that control the smart device to execute operations.

[0114] Table 4: Interaction Information Table (3)

[0115] time Interaction Intent Scene Wake-up event 2020.12.01 12:30:05 none none yes 2020.12.01 12:30:25 Play children's content child no 2021.01.30 12:54:08 none none yes 2021.01.30 12:54:23 Play children's content child no 2021.02.20 11:00:01 none none yes 2021.02.20 11:00:15 Play music entertainment no

[0116] S406, Receive first voice. Here, the first voice refers to the user's voice containing the wake word when waking up the smart device again.

[0117] Taking the second voice prompt as "play the music from the party" as an example, after executing step S404 above, the smart device starts playing the party song. If the user attempts to wake up the smart device again while it is playing the party song, the user needs to speak the voice prompt containing the wake-up word again. In this embodiment of the invention, the voice prompt containing the wake-up word spoken by the user again is recorded as the "first voice prompt." The microphone 201 of the smart device collects the first voice prompt input by the user, for example, "Xiaoyi Tongxue."

[0118] S408, determine the confidence level of the wake-up word contained in the first speech; query the previous round of interaction intent corresponding to the first speech (i.e., the current round of interaction intent corresponding to the second speech); determine the first scenario and the wake-up threshold corresponding to the first scenario based on the previous round of interaction intent corresponding to the first speech.

[0119] The wake-up engine 206 queries the sensitivity control module 207 to obtain the interaction intent of the previous round corresponding to the first voice (i.e., the interaction intent of the current round corresponding to the second voice), and then determines the first scene and the wake-up threshold corresponding to the first scene according to the intent-scene mapping table (i.e., Table 1).

[0120] First, taking "Xiaoyi Tongxue" as the first speech as an example, the process by which the wake-up engine 206 determines the confidence level of the wake-up word contained in the first speech is described in detail:

[0121] The audio signal acquisition and processing module 202 converts the first speech into an audio stream and sends it to the audio stream RingBuffer 203. The wake-up engine 206 monitors the audio stream in the Ring Buffer 203 in real time and segments it into several speech segments. For example, the first speech "Xiaoyi Tongxue" is segmented into three speech segments: "Xiaoyi" (speech segment M), "Yitong" (speech segment N), and "Tongxue" (speech segment P). The wake-up engine 206 performs streaming processing on speech segments M, N, and P to obtain the confidence level of the wake-up words contained in the first speech. Then, the interaction behavior statistics submodule 2071 of the sensitivity control module 207 records the first wake-up event corresponding to the first speech. Finally, the sensitivity control module 207 calculates that the time difference between the first wake-up event corresponding to the first speech recorded by the interaction behavior statistics submodule 2071 and the time of the interaction intent in this round corresponding to the second speech recorded by the submodule 2071 is less than a preset time threshold. Therefore, the sensitivity control module 207 determines that the first speech and the second speech are two user voices in the same session.

[0122] The following example illustrates how the wake-up engine 206 retrieves the wake-up threshold corresponding to the first scenario, "Entertainment," from the sensitivity control module 207, using the second voice prompt as "Play songs for the party," the first scenario as "Entertainment," and the first voice prompt as "Xiaoyi Tongxue."

[0123] First, in response to the wake-up engine 206's request to obtain the wake-up threshold, the sensitivity control module 207 queries the interaction information table and finds that the previous interaction intent in the same session is "play music," which is also the interaction intent corresponding to the second voice mentioned above. Then, the sensitivity control module 207 determines, according to the intent-scene mapping table (i.e., Table 1), that the first scene corresponding to the previous interaction intent "play music" in the same session is "entertainment," and the wake-up threshold corresponding to the first scene "entertainment" is 0.3837. Next, the sensitivity control module 207 returns the wake-up threshold of 0.3837 to the wake-up engine 206, so that the wake-up engine 206 queries the sensitivity control module 207 to obtain the wake-up threshold of 0.3837 corresponding to the first scene.

[0124] S410, determine whether to wake up the smart device based on the confidence level of the wake-up word contained in the first voice and the wake-up threshold corresponding to the first scenario.

[0125] Taking the first scenario as "entertainment" and the first voice as "Xiaoyi Tongxue" as an example, step S412 is described in detail: When the wake-up engine 206 processes the voice segment M ("Xiaoyi") obtained in S410, if the confidence of the output wake-up word is 0.5, which is greater than the wake-up threshold of 0.3837 corresponding to the first scenario "entertainment", then the wake-up engine 206 stops processing the subsequent voice segments and wakes up the smart device again.

[0126] S412, If the device is woken up, update the wake-up probability of the smart device in the first scenario.

[0127] After a smart device is woken up, before updating the wake-up probability of the smart device in the first scenario, it is necessary to first update the number of "scene triggers" and the number of "scene-wake-up event pairs" corresponding to the first scenario. The following describes in detail the process of updating the number of "scene triggers" and the number of "scene-wake-up event pairs" corresponding to the first scenario, still using "entertainment" as the first scenario as an example:

[0128] First, the process of updating the number of "scene triggers" n(c1) corresponding to the first scene "entertainment" (denoted as c1) is described in detail, taking the second voice as "play music for the party" and the first scene corresponding to the second voice as "entertainment" as an example.

[0129] Assume the original value of n(c1) is 299, meaning that the first scenario "Entertainment" was triggered a total of 299 times before this scenario was triggered. After this scenario was triggered, the number of times the first scenario "Entertainment" was triggered increased by one. Therefore, n(c1) is updated to n(c1) = 299 + 1 = 300. The original value of the number of "scenario-wake-up event pairs" n(c1, w) corresponding to the first scenario "Entertainment" is 200, meaning that the smart device was woken up a total of 200 times within the preset time threshold under the first scenario "Entertainment".

[0130] It should be noted that n(c1)-n(c1,w)=300-200=100 means that when the "entertainment" scene is triggered 300 times, the cumulative number of times the smart device is not woken up within the preset time threshold for each "entertainment" scene trigger is 100 times.

[0131] Secondly, the process of updating the number of "scene-awakening event pairs" n(c1) corresponding to the first scene (called c1) of "entertainment" is described in detail, taking the wake-up event that occurs in the first scene of "entertainment" as an example.

[0132] After determining the confidence level of the wake-up word contained in the first speech in S408, the wake-up engine 206 sends the wake-up event "previous round's interaction intent - play music, wake up" (referred to as the first wake-up event) to the sensitivity control module 207 for recording. The interaction behavior statistics submodule 2071 of the sensitivity control module 207 writes the first wake-up event and the time of receiving the first wake-up event into the interaction information table, that is, when a new wake-up event occurs, Table 4 will be updated. The interaction behavior statistics submodule 2071 adds the record of the first wake-up event to Table 4, and the updated Table 4 is shown in Table 5. The record of the first wake-up event includes the time of receiving the first wake-up event and whether it is a wake-up event. The record of the first wake-up event is shown in the last row of data in Table 5. Table 5 has an additional first wake-up event at 11:01:15 on February 20, 2021 compared to Table 4. If the sensitivity control module 207 determines that the time difference between the first moment of the first wake-up event recorded by the interaction behavior statistics submodule 2071 and the second moment of the "previous round of interaction intent - playing music" recorded by the interaction behavior statistics submodule 2071 is less than a preset time threshold, it means that the first voice and the second voice are in the same conversation. At this time, the number of "scene-wake-up event pairs" n(c1,w) corresponding to the first scene "entertainment" is incremented by 1.

[0133] In this embodiment of the invention, it is assumed that the preset time threshold is 3 minutes. For example, the wake-up engine 206 sends the first wake-up event, "the previous interaction intent - play music, wake up," to the sensitivity control module 207 at 11:01:15 on February 20, 2021. The sensitivity control module 207 determines that the time difference between the first moment recorded by the interaction behavior statistics submodule 2071 for the first wake-up event (11:01:15 on February 20, 2021) and the second moment recorded by the interaction behavior statistics submodule 2071 for "the previous interaction intent - play music" (11:00:15 on February 20, 2021) is 1 minute, which is less than the preset time threshold of 3 minutes. Therefore, the first voice and the second voice are in the same conversation. The scene recognition submodule 2072 of the sensitivity control module 207 maps the previous interaction intent "play music" to the first scene "entertainment," and updates the number of "scene-wake-up event pairs" n(c1,w) corresponding to the first scene "entertainment." That is, after the first wake-up event occurs, the number of times the smart device is woken up within the preset time threshold in the "entertainment" scene increases by one. Therefore, n(c1,w) is updated to n(c1,w) = 200 + 1 = 201.

[0134] Then, based on the updated number of "scene triggers" and "scene-wake-up event pairs," the wake-up probability of the smart device in the first scene can be updated. Taking "entertainment" as the first scene as an example, we have:

[0135] According to formula (1), when i = 1, the wake-up probability of the smart device in the first scenario of "entertainment" is:

[0136]

[0137] Here, c1 represents the first scenario as "entertainment", and w represents the first wake-up event.

[0138] S414, Update the wake-up threshold corresponding to the first scenario based on the updated wake-up probability of the smart device in the first scenario.

[0139] According to formula (2) for the wake-up threshold corresponding to a certain scenario:

[0140]

[0141] Wherein, f(P(w|c) i ))=(P(w|c i )-P(w|c0))α, Let P(w|c0) and P(w|c0) represent the wake-up threshold and the wake-up probability of the smart device, respectively, for the "daily standby" scenario, and α be the adjustment range. Then, the wake-up threshold for the first scenario, "entertainment," is:

[0142]

[0143] In this embodiment of the invention, α is assumed to be 0.7. It can be seen that the wake-up threshold corresponding to the first scenario of "entertainment" decreases from 0.3837 to 0.38135. Due to the decrease in the wake-up threshold, the wake-up sensitivity will improve. As can be seen from the above, based on the previous interaction intent of "play music," the wake-up threshold is automatically adjusted, thereby automatically adjusting the wake-up sensitivity. The smart device responds to the user's second voice and begins playing music. At this time, due to the music playing, the ambient noise around the smart device becomes louder than before the music started playing, making it more difficult for the smart device to capture the user's voice command. Therefore, theoretically, the smart device needs to lower the wake-up threshold at this time, that is, increase the wake-up sensitivity, so that the smart device can more sensitively detect the user's voice containing the wake word. This embodiment of the invention provides... Figure 4 The method flow shown is as follows: This method automatically identifies the scenario based on the user's interaction intent in the previous round in the same session; automatically calculates the wake-up probability of the smart device in the scenario based on historical interaction behavior data; and automatically adjusts the wake-up threshold based on the wake-up probability of the smart device in the scenario, thereby automatically adjusting the wake-up sensitivity.

[0144] It should be noted that if the time difference between the first moment of the first wake-up event recorded by the interaction behavior statistics submodule 2071 and the second moment of "the previous round of interaction intent - playing music" recorded by the interaction behavior statistics submodule 2071 is greater than the preset time threshold of 3 minutes, the sensitivity control module 207 determines that the first voice and the second voice are not in the same conversation, and thus determines that the smart device was woken up in the "daily standby" scenario.

[0145] Table 5: Interaction Information Table (4)

[0146] time Interaction Intent Scene Wake-up event 2020.12.01 12:30:05 none none yes 2020.12.01 12:30:25 Play children's content child no 2021.01.30 12:54:08 none none yes 2021.01.30 12:54:23 Play children's content child no 2021.02.20 11:00:01 none none yes 2021.02.20 11:00:15 Play music entertainment no 2021.02.20 11:01:15 none none yes

[0147] Repeat the above steps to continuously update the wake-up probability P(w|c) of the smart device. i This continuously updates the wake-up threshold. This allows for continuous adjustment of the wake-up sensitivity.

[0148] Figure 5 Another flowchart of a smart device wake-up method provided in an embodiment of the present invention includes: S502-S508;

[0149] S502, Receive a first voice message containing a wake-up word, the wake-up word being a pre-set word used to wake up the smart device.

[0150] In this embodiment of the invention, a pre-set word for waking up the smart device is used as the wake-up word. The smart device can receive a first voice message containing the wake-up word.

[0151] S504, obtain the confidence level of the wake word contained in the first speech.

[0152] In an embodiment of the present invention, after receiving a first voice containing a wake-up word, the smart device obtains the confidence level of the wake-up word contained in the first voice.

[0153] S506, based on the interaction intent of the previous round corresponding to the first voice, determine the first scenario and the wake-up threshold corresponding to the first scenario; wherein, the interaction intent of the previous round is the interaction intent corresponding to the second voice, the second voice is the voice received before receiving the first voice, the second voice contains an instruction statement for controlling the smart device to perform an operation, and the first scenario is the scenario corresponding to the interaction intent of the previous round.

[0154] In this embodiment of the invention, the wake-up threshold in this step includes a threshold that is automatically updated based on historical interaction behavior data. The historical interaction behavior data here includes the number of "scene triggers" and the number of "scene-wake-up event pairs". The number of "scene triggers" here is the cumulative number of triggers in a certain scene, and the number of "scene-wake-up event pairs" here is the cumulative number of wake-ups of the smart device in a certain scene.

[0155] S508, if the confidence level is greater than the wake-up threshold, wake up the smart device.

[0156] The above technical solution enables the determination of the corresponding scenario and wake-up threshold based on the interaction intent of the previous round, thereby determining whether to wake up the smart device based on the confidence level and the wake-up threshold.

[0157] In this embodiment of the invention, firstly, when the smart device is woken up, the number of "scene triggers" and the number of "scene-wake-up event pairs" corresponding to the first scenario are updated. Secondly, based on the updated number of "scene triggers" and the number of "scene-wake-up event pairs" corresponding to the first scenario, the wake-up probability corresponding to the first scenario is updated. Specifically, the wake-up probability corresponding to the first scenario is updated according to the following formula:

[0158]

[0159] Wherein, P(w|c1) represents the wake-up probability corresponding to the first scenario, n(c1,w) represents the number of "scenario-wake-up event pairs" corresponding to the first scenario, and n(c1) represents the number of "scenario triggers" corresponding to the first scenario. Finally, the wake-up threshold corresponding to the first scenario is updated according to the updated wake-up probability of the smart device corresponding to the first scenario. Specifically, the wake-up threshold corresponding to the first scenario is updated according to the following formula:

[0160]

[0161] Among them, the Used to represent the wake-up threshold corresponding to the first scenario, The following parameters are used to represent the wake-up threshold corresponding to the daily standby scenario, where the wake-up threshold for the daily standby scenario is 0.5. P(w|c1) represents the wake-up probability corresponding to the first scenario, P(w|c0) represents the wake-up probability corresponding to the daily standby scenario, and α represents the adjustment range. As can be seen from the above technical solution, this embodiment of the invention first automatically adjusts the wake-up probability corresponding to the first scenario based on historical interaction behavior data, and then automatically adjusts the wake-up threshold corresponding to the first scenario based on the adjusted wake-up probability.

[0162] It should be noted that, in this embodiment of the invention, the interaction behavior statistics submodule 2071 records the first wake-up event corresponding to the first voice and determines the first moment of the first wake-up event corresponding to the first voice. Further, the sensitivity control module 207 determines the time difference between the first moment and the second moment, wherein the second moment is the moment of the interaction intent corresponding to the second voice as determined by the interaction behavior statistics submodule 2071; if the time difference is less than a preset time threshold, the sensitivity control module 207 determines that the first voice and the second voice are in the same session.

[0163] This invention also provides a smart device wake-up device, including at least one processor, the processor being configured to execute a program stored in a memory, wherein when the program is executed, the device performs the following steps:

[0164] The system receives a first voice message containing a wake-up word, which is a pre-set word used to wake up the smart device; obtains the confidence level of the wake-up word contained in the first voice message; determines a first scenario and a wake-up threshold corresponding to the first scenario based on the previous round of interaction intent corresponding to the first voice message; wherein the previous round of interaction intent is the interaction intent corresponding to a second voice message, the second voice message is the voice message received before receiving the first voice message, the second voice message contains an instruction statement for controlling the smart device to perform an operation, and the first scenario is the scenario corresponding to the previous round of interaction intent; and wakes up the smart device when the confidence level is greater than the wake-up threshold.

[0165] This invention also provides a computer program product containing instructions that, when run on a computer, cause the following steps to be executed by the computer:

[0166] The system receives a first voice message containing a wake-up word, which is a pre-set word used to wake up the smart device; obtains the confidence level of the wake-up word contained in the first voice message; determines a first scenario and a wake-up threshold corresponding to the first scenario based on the previous round of interaction intent corresponding to the first voice message; wherein the previous round of interaction intent is the interaction intent corresponding to a second voice message, the second voice message is the voice message received before receiving the first voice message, the second voice message contains an instruction statement for controlling the smart device to perform an operation, and the first scenario is the scenario corresponding to the previous round of interaction intent; and wakes up the smart device when the confidence level is greater than the wake-up threshold.

[0167] This invention also provides a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the following steps are performed:

[0168] The system receives a first voice message containing a wake-up word, which is a pre-set word used to wake up the smart device; obtains the confidence level of the wake-up word contained in the first voice message; determines a first scenario and a wake-up threshold corresponding to the first scenario based on the previous round of interaction intent corresponding to the first voice message; wherein the previous round of interaction intent is the interaction intent corresponding to a second voice message, the second voice message is the voice message received before receiving the first voice message, the second voice message contains an instruction statement for controlling the smart device to perform an operation, and the first scenario is the scenario corresponding to the previous round of interaction intent; and wakes up the smart device when the confidence level is greater than the wake-up threshold.

[0169] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0170] The above descriptions are merely a few embodiments of the present invention. Those skilled in the art can make various modifications or variations to the present invention based on the information disclosed in the application documents without departing from the spirit and scope of the present invention.

Claims

1. A method for waking up a smart device, applied to a smart device, characterized in that, The method includes: Receive a first voice message containing a wake-up word, which is a pre-set word used to wake up the smart device; Obtain the confidence level of the wake word contained in the first speech; Based on a preset time threshold, it is determined whether the first voice and the second voice are in the same session. The second voice is the voice received before the first voice is received. The second voice contains instructional statements for controlling the smart device to perform operations. When the first voice and the second voice are in the same session: Based on the interaction intent of the previous round corresponding to the first voice, a first scenario and a wake-up threshold corresponding to the first scenario are determined; wherein, the interaction intent of the previous round is the interaction intent corresponding to the second voice, and the first scenario is the scenario corresponding to the interaction intent of the previous round. If the confidence level is greater than the wake-up threshold, the smart device is woken up, and the wake-up threshold corresponding to the first scenario is reduced based on the wake-up probability corresponding to the first scenario and the wake-up probability corresponding to the daily scenario; wherein, the wake-up probability corresponding to any scenario is determined based on the number of "scenario triggers" and the number of "scenario-wake-up event pairs" corresponding to that scenario, the number of "scenario triggers" is the cumulative number of triggers in a certain scenario, and the number of "scenario-wake-up event pairs" is the cumulative number of wake-ups of the smart device in a certain scenario; If the confidence level is less than the wake-up threshold, the wake-up threshold corresponding to the first scenario is increased based on the wake-up probability corresponding to the first scenario and the wake-up probability corresponding to the daily scenario.

2. The method according to claim 1, characterized in that, The method further includes: When the smart device is woken up, update the number of "scene triggers" and the number of "scene-wake-up event pairs" corresponding to the first scenario; Based on the updated number of "scene triggers" and the number of "scene-wake-up event pairs" corresponding to the first scenario, update the wake-up threshold corresponding to the first scenario.

3. The method according to claim 2, characterized in that, The step of updating the wake-up threshold corresponding to the first scenario based on the updated number of "scenario triggers" and the number of "scenario-wake-up event pairs" for the first scenario specifically includes: Update the wake-up probability corresponding to the first scenario based on the updated number of "scenario triggers" and the number of "scenario-wake-up event pairs" corresponding to the first scenario; The wake-up threshold corresponding to the first scenario is updated based on the updated wake-up probability of the smart device corresponding to the first scenario.

4. The method according to claim 3, characterized in that, The wake-up probability corresponding to the first scenario is calculated based on the following formula: Wherein, P(w|c1) represents the wake-up probability corresponding to the first scenario, n(c1,w) represents the number of "scenario-wake-up event pairs" corresponding to the first scenario, and n(c1) represents the number of "scenario triggers" corresponding to the first scenario.

5. The method according to claim 4, characterized in that, The wake-up threshold for the first scenario is calculated based on the following formula: Among them, the Used to represent the wake-up threshold corresponding to the first scenario, The values ​​are used to represent the wake-up threshold corresponding to the daily standby scenario, where the wake-up threshold for the daily standby scenario is 0.

5. P(w|c1) represents the wake-up probability corresponding to the first scenario, P(w|c0) represents the wake-up probability corresponding to the daily standby scenario, and α represents the adjustment range.

6. The method according to any one of claims 1 to 5, characterized in that, The method further includes: Record the first wake-up event corresponding to the first voice, and determine the first moment of the first wake-up event corresponding to the first voice.

7. The method according to claim 6, characterized in that, The method further includes: Determine the time difference between the first moment and the second moment, wherein the second moment is the moment when the interactive intent corresponding to the second voice is determined; If the time difference is less than a preset time threshold, it is determined that the first voice and the second voice are in the same session.

8. A smart device wake-up device, characterized in that, It includes at least one processor for executing a program stored in a memory, which, when executed, causes the device to perform the method as described in any one of claims 1-7.

9. A computer program product containing instructions, characterized in that, When the computer program product is run on a computer, the method described in any one of claims 1-7 is executed by the computer.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed by a processor, performs the method as described in any one of claims 1-7.

Images