Remote control method and system for household intelligent sound equipment

Abstract

The invention discloses a remote control method and system for household intelligent sound equipment, and belongs to the technical field of Internet of Things. The method comprises the following steps: constructing a static space topology model; each intelligent sound equipment acquires respective voice signals in parallel and synchronously acquires respective binary staring state marks; executing a first-level arbitration of line-of-sight priority; if the first-stage arbitration does not generate the final response unit, starting second-stage arbitration; when it is judged that no intelligent sound equipment is watched, returning to position instruction analysis based on the static space topology model; and the final response unit executes the voice instruction. According to the invention, by constructing the spatial topology model, implementing a dual arbitration mechanism of sight line detection and sound source localization and state synchronous broadcasting, the problems of equipment response disorder, resource waste and user experience splitting caused by uncontrollable instruction propagation range in the prior art are solved.

Description

Technical Field

[0001] This invention relates to the field of Internet of Things (IoT) technology, and in particular to a remote control method and system for a home smart speaker. Background Technology

[0002] With the rapid development of artificial intelligence and the Internet of Things (IoT) technologies, smart home systems are gradually becoming widespread in various aspects of family life. As one of the core interactive devices in a smart home system, home smart speakers not only provide a high-quality music playback experience but also allow users to control other smart devices via voice commands, bringing a more convenient and comfortable living environment.

[0003] Existing multi-room smart speaker systems achieve whole-house audio coverage and voice control by deploying multiple speaker devices in a distributed manner and using a unified cloud platform or local network for device management.

[0004] In existing technologies, multi-room audio systems typically employ simple signal strength detection or basic voice recognition to determine the responding device, resulting in uncontrollable command propagation range. This issue further leads to cross-device response conflicts and inconsistent states, with audio devices in different locations potentially executing contradictory operating commands, disrupting the consistency of the user experience. Simultaneously, resource competition and fragmented user experiences among multiple devices force users to manually intervene in the operation of each device, severely weakening the automation advantages of intelligent systems and limiting the practical effectiveness of multi-room intelligent audio systems. Summary of the Invention

[0005] This application provides a remote control method and system for home smart speakers, which solves the problems of response conflicts, resource waste, and fragmented user experience caused by the uncontrollable command propagation range in existing multi-room audio systems. It achieves the technical effects of precise directional response to target devices, conflict-free collaborative operation between multiple devices, and natural and efficient interaction between users and the smart speaker system.

[0006] This application provides a remote control method for a home smart speaker, including: pre-configuring and storing the binding relationship between each smart speaker device and a logical location identifier, and constructing a static spatial topology model; When a user issues a voice command, each smart speaker device collects its own voice signal in parallel and simultaneously acquires its own binary gaze status marker. It aggregates the gaze status tags of all smart speaker devices and performs the first-level arbitration based on gaze priority: if it determines that there is a unique gazed device, it is directly identified as the final response unit; If the first-level arbitration does not produce a final response unit, the second-level arbitration is initiated: when it is determined that there are multiple smart speaker devices being monitored, sound source localization arbitration is only performed between the monitored devices. If it is determined that there is no smart speaker device being observed, the process reverts to the position command parsing based on the static spatial topology model to determine the final response unit. The final response unit executes the voice command and broadcasts a status synchronization message to other devices.

[0007] Furthermore, the steps for constructing a static spatial topology model include: Collect three-dimensional spatial layout data of the home environment and generate a digital map using a laser scanner; Assign a unique logical location identifier to each smart speaker device, and bind the physical coordinates of each smart speaker device to the logical identifier; Based on the relative distance and orientation between each smart speaker device, a topological graph is established, where nodes represent devices and edges represent spatial connectivity and distance weights. Store the binding relationship and topology diagram between the physical coordinates and logical identifiers of each smart speaker device in the database.

[0008] Furthermore, the steps for synchronously acquiring their respective binary observed state tags include: Get in the time window Internally acquired sensor signals ; Calculate each smart speaker device Weighted integral perception intensity and the preset gaze threshold The binary observed state marker is generated through comparison. ; Right now: ; ; In the formula, For device indexing, Indicates being watched. Indicates that you are not being watched. For equipment The weighted integral perceived intensity, To detect the start time, For the detection time window, For equipment The total number of sensing elements, For the first Each sensing element in Instantaneous signal reading at time 10:00. For the first The environmental baseline signal of each sensing element For the first Preset weighting coefficients for each sensing element The calibrated gaze judgment threshold.

[0009] Furthermore, the steps for arbitrating sound source localization among the monitored devices include: Define the set of multiple observed smart speaker devices as the first candidate set. ; First candidate set Each smart speaker device Acquire the collected voice signal Calculate its location at all possible sound source locations. The steering response power is obtained by applying a phase transformation for weighting. : ; Each smart speaker device Calculate its maximum response power: ; Based on the first candidate set In this study, the smart speaker with the highest response power was selected as the final response device. : ; In the formula, For equipment The The signal from each microphone in the frequency domain Indicates complex conjugation. Angular frequency, Let be the sound source location vector in three-dimensional space. For the sound source in Position to number The propagation delay of each microphone, The imaginary unit, It is a natural constant.

[0010] Furthermore, the step of reverting to the position command parsing based on the static spatial topology model includes: Perform natural language processing on voice commands to extract location keywords; The extracted location keywords are matched with logical location identifiers in the static spatial topology model, and a similarity score is calculated. If a unique smart speaker device is matched, it will be selected as the final responding smart speaker device. If multiple smart speaker devices are matched, the system will further combine sound source localization-assisted arbitration to select the device closest to the sound source as the final response device. If no specific location keyword is specified, the default main device in the topology model will be selected.

[0011] Furthermore, the steps for calculating the similarity score include: The extracted location keywords are stemmed and normalized to generate standardized keyword vectors; Standardized keywords and logical position identifiers are converted into high-dimensional vector representations using a pre-trained word embedding model; Calculate the cosine similarity between the keyword vector and each logical position identifier vector, and use this as the similarity score. : ; In the formula, This is a vector representation of the keywords. A vector representation of a logical position identifier; If the similarity score is not less than the preset similarity threshold, the corresponding smart speaker device will be included in the candidate list. The device corresponding to the logical location name with the highest similarity score is selected as the candidate response smart speaker device. If multiple smart speaker devices have the same score and it is not less than the similarity threshold, then sound source localization arbitration is initiated among these devices.

[0012] Furthermore, in conjunction with sound source localization-assisted arbitration, the steps for selecting the device closest to the sound source as the final response device include: Each smart speaker device collects voice signals and calculates the three-dimensional coordinates of the sound source in space based on a time difference of arrival (TDOA) localization algorithm. ; Based on the device location coordinates stored in the static spatial topology model Calculate the Euclidean distance between each device and the sound source: ; Choose the device closest to the sound source as the final response device, that is: ; In the formula, For the estimated sound source location coordinates, For equipment The preset position coordinates For equipment The distance between the sound source and the sound source.

[0013] This application provides a remote control system for a home smart speaker, which implements a remote control method for a home smart speaker, including: a model building module, a tag acquisition module, a first-level arbitration execution module, a second-level arbitration execution module, a rollback module, and an instruction execution module; The model building module is used to pre-configure and store the binding relationship between each smart speaker device and a logical location identifier, and to build a static spatial topology model. The tag acquisition module is used to allow each smart speaker device to collect its own voice signal in parallel and simultaneously acquire its own binary gazed state tag when the user issues a voice command. The first-level arbitration execution module is used to aggregate the gaze status markers of all smart speaker devices and execute the first-level arbitration with gaze priority: if it is determined that there is a unique gazed device, it is directly identified as the final response unit; The second-level arbitration execution module is used to initiate the second-level arbitration if the first-level arbitration does not produce a final response unit: when it is determined that there are multiple smart speaker devices being monitored, sound source localization arbitration is only performed between the monitored devices; The rollback module is used to roll back to the position command parsing based on the static spatial topology model to determine the final response unit when it is determined that there is no smart speaker device being watched. The instruction execution module is used to execute voice instructions by the final response unit and broadcast status synchronization messages to other devices.

[0014] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages: By clearly defining the logical location and control range of each speaker device, the problem of uncontrollable command propagation range is effectively solved, avoiding resource waste and auditory interference caused by multiple speakers receiving and executing voice commands simultaneously throughout the home. Furthermore, by introducing a first-level arbitration mechanism prioritizing line-of-sight, the system can intelligently identify the device the user actually wants to control. Even when multiple devices receive commands simultaneously, it can accurately select the device the user is looking at as the response unit, fundamentally eliminating the problem of conflicting responses from speakers in different locations to the same command and ensuring consistency in the status of each device. Going further, through a mechanism where the final response unit broadcasts status synchronization messages to other devices, the operating status of all devices can be updated in a timely manner, effectively preventing resource competition such as multiple devices vying for the same audio source. This avoids the tedious manual intervention required by users for each device, thus ensuring a complete automated experience for the smart home system and improving user convenience and satisfaction. Attached Figure Description

[0015] Figure 1A flowchart illustrating a remote control method for a home smart speaker provided in this application embodiment; Figure 2 This is a schematic diagram of the structure of a remote control system for a home smart speaker provided in an embodiment of this application. Detailed Implementation

[0016] This application provides a remote control method and system for home smart speakers, which solves the problems of command response conflicts and resource competition in multi-device environments in the prior art. By constructing a spatial topology model, implementing a dual arbitration mechanism of line-of-sight priority and sound source localization, and synchronous broadcasting of status, it achieves accurate identification and collaborative control of target devices, effectively improving the system's response accuracy and user experience.

[0017] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0018] like Figure 1 The diagram shown is a flowchart of a remote control method for a home smart speaker provided in an embodiment of this application. The method is applied to a remote control system for a home smart speaker and includes the following steps: pre-configuring and storing the binding relationship between each smart speaker device and a logical location identifier, and constructing a static spatial topology model. The binding relationship between each smart speaker device and the logical location identifier is obtained as follows: Users use the configuration application on their smartphones or tablets to scan the QR code or NFC tag on each smart speaker device to obtain the device's unique physical identifier (MAC address or device serial number). The application displays a home floor plan or a list of rooms. Users can manually drag and drop each smart speaker device to its corresponding room location, which automatically generates a logical location identifier in the format of "room name + device type + serial number", such as "living room - speaker - 01". The binding relationship is uploaded to the cloud server for storage via Wi-Fi, and users can modify and update the binding relationship at any time through the application.

[0019] When a user issues a voice command, each smart speaker device collects its own voice signal in parallel and simultaneously activates its gaze detection sensor to obtain its own binary gaze status marker. The central coordination server gathers the gaze status tags of all smart speaker devices and performs the first-level arbitration based on gaze priority: if it determines that there is a unique gazed device, it is directly identified as the final response unit; The central coordination server is a cloud server cluster, which is a remote server provided by the smart speaker manufacturer. It maintains a long-term connection with all smart speaker devices via the Internet and is responsible for arbitration calculations and status synchronization.

[0020] If the first-level arbitration does not produce a final response unit, the second-level arbitration is initiated: when it is determined that there are multiple smart speaker devices being monitored, sound source localization arbitration is only performed between the monitored devices. If it is determined that there is no smart speaker device being observed, the process reverts to the position command parsing based on the static spatial topology model to determine the final response unit. The final response unit executes voice commands and broadcasts status synchronization messages to other devices to update their status log tables and prevent conflict responses.

[0021] Furthermore, the steps for constructing a static spatial topology model include: The three-dimensional spatial layout data of the home environment is collected by the built-in LiDAR scanner of the mobile phone, including room division, wall position and furniture distribution, and a digital map is generated by the LiDAR scanner. Assign a unique logical location identifier to each smart speaker device, such as "main living room speaker" or "auxiliary kitchen speaker", and bind the physical coordinates of each smart speaker device to the logical identifier; Based on the relative distance and orientation between each smart speaker device, a topological graph is established, where nodes represent devices and edges represent spatial connectivity and distance weights. The topology graph is constructed using graph theory algorithms in IoT networks. The specific steps are as follows: First, the signal strength (RSSI), distance, and relative orientation between smart speaker devices are measured using their built-in Wi-Fi or Bluetooth modules. Second, the devices are treated as nodes in the topology graph, and edges are constructed based on the measured distances and relative orientations, where edge weights represent spatial connectivity and Euclidean distance. Then, Dijkstra's algorithm is applied to optimize the graph structure, ensuring connectivity and minimizing path costs. Finally, the topology graph is stored in a database on a central coordination server. This method is based on existing IoT topology optimization techniques, such as mesh topology construction in the ZigBee protocol.

[0022] The binding relationship and topology diagram between the physical coordinates and logical identifiers of each smart speaker device are stored in the database of the central coordination server, which supports subsequent location command parsing and inter-device collaboration.

[0023] Furthermore, the steps for synchronously acquiring their respective binary observed state tags include: Through line-of-sight detection sensors (including) (a low-power infrared sensing element) acquires data within the time window. Internally acquired sensor signals ; Calculate each smart speaker device Weighted integral perception intensity and the preset gaze threshold The binary observed state marker is generated through comparison. ; Right now: ; ; In the formula, For device indexing, Indicates being watched. Indicates that you are not being watched. For equipment The weighted integral perceived intensity, To detect the start time, To detect the time window, timing is performed using the device's system clock chip. For equipment The total number of sensing elements, that is, the number of sensors determined during the hardware design of the device. For the first Each sensing element in The instantaneous signal reading at any given moment is obtained through an infrared photodiode sensor integrated into the smart speaker device. This sensor converts the received infrared light intensity into a voltage signal. For the first The environmental baseline signal of each sensing element is obtained through baseline measurement using the same infrared sensor in a user-free state. For the first The preset weighting coefficients of each sensing element are written into the device's firmware parameters during sensor factory calibration. These parameters characterize the contribution of elements at different positions to the orientation of the face. The calibrated gaze judgment threshold is an empirical threshold determined through statistical analysis of a large amount of user gaze data collected in a laboratory environment, and is written into the device firmware.

[0024] Furthermore, the steps for arbitrating sound source localization among the monitored devices include: Define multiple smart speaker devices being watched (i.e.) The set of ) is the first candidate set ; First candidate set Each smart speaker device Utilizing its microphone array (including Each microphone (determined by the device hardware specifications) acquires the collected voice signal. Calculate its location at all possible sound source locations. The steering response power is obtained by applying a phase transformation for weighting. : ; Each smart speaker device Calculate its maximum response power: ; The central coordination server is based on the first candidate set. In this study, the smart speaker with the highest response power was selected as the final response device. : ; In the formula, For equipment The The frequency domain signal from each microphone is acquired through an analog audio signal collected by a MEMS microphone array integrated into the smart speaker device. This analog signal is then converted into a digital signal by an ADC (Analog-to-Digital Converter), and finally, the frequency domain signal is obtained by an FFT (Functional-to-Flight Transform) performed by a DSP (Digital Signal Processor). Indicates complex conjugation. The angular frequency is determined by the signal sampling rate. Given the sound source location vector in three-dimensional space, all possible three-dimensional spatial location vectors are enumerated by scanning an array. For the sound source in Position to number The propagation delay of each microphone is obtained through the geometric positional relationship of the microphone array and the speed of sound propagation. The position of the microphone array is determined during equipment manufacturing. The imaginary unit, It is a natural constant.

[0025] Furthermore, the step of reverting to the position command parsing based on the static spatial topology model includes: Natural language processing is performed on voice commands to extract location keywords, such as living room, kitchen, or bedroom; The extracted location keywords are matched with logical location identifiers in the static spatial topology model, and a similarity score is calculated. If the highest similarity score is not less than the preset similarity threshold, and the device with that score is unique, then the device is determined to be a unique smart speaker device.

[0026] If multiple devices have the same similarity score and none of them are less than the preset similarity threshold, then sound source localization arbitration will be initiated among these multiple devices.

[0027] If all scores are less than the preset similarity threshold, then the preset master device or the nearest device in the topology model is selected.

[0028] If a unique smart speaker device is matched, it will be selected as the final responding smart speaker device. If multiple smart speaker devices are matched, the system will further combine sound source localization-assisted arbitration to select the device closest to the sound source as the final response device. If no specific location keyword is specified, the default main device in the topology model will be selected.

[0029] Furthermore, the steps for calculating the similarity score include: The extracted location keywords are stemmed and normalized to generate standardized keyword vectors; Standardized keywords and logical position identifiers are converted into high-dimensional vector representations using pre-trained word embedding models (such as Word2Vec or BERT). The steps for building a word embedding model are as follows: Training set: Uses large-scale Chinese corpora, such as: Baidu Encyclopedia full-text data (more than 10 million entries), Chinese Wikipedia dump data.

[0030] Training methods: Word2Vec: using Skip-gram or CBOW models, trained on the TensorFlow framework, or BERT: using a pre-trained Chinese model (such as the Chinese BERT model from Harbin Institute of Technology) for domain adaptation training.

[0031] Output set: a 300-dimensional word vector space, where each word is mapped to a point in the space.

[0032] Similarity calculation: Cosine similarity is used to measure semantic distance.

[0033] Calculate the cosine similarity between the keyword vector and each logical position identifier vector, and use this as the similarity score. : ; In the formula, The vector representation of the keywords is calculated by the NLP processing module on the cloud server, which loads a pre-trained word embedding model. The vector representation of the logical position identifier is obtained by vectorizing the logical position identifier through the same word embedding model; If the similarity score is not less than the preset similarity threshold, the corresponding smart speaker device will be included in the candidate list. The similarity threshold is obtained through an experimental optimization process, which involves adjusting the threshold on the test dataset to maximize matching accuracy and minimize false positives, for example, by setting the threshold to 0.8 using cross-validation.

[0034] The device corresponding to the logical location name with the highest similarity score is selected as the candidate response smart speaker device. If multiple smart speaker devices have the same score and it is not less than the similarity threshold, then sound source localization arbitration is initiated among these devices.

[0035] Furthermore, in conjunction with sound source localization-assisted arbitration, the steps for selecting the device closest to the sound source as the final response device include: Each smart speaker device uses a microphone array to collect voice signals and calculates the three-dimensional coordinates of the sound source in space based on a time difference of arrival (TDOA) localization algorithm. ; Based on the device location coordinates stored in the static spatial topology model Calculate the Euclidean distance between each device and the sound source: ; Choose the device closest to the sound source as the final response device, that is: ; In the formula, The estimated sound source location coordinates are obtained from audio signals collected by the microphone arrays of each smart speaker device, based on the TDOA (Time Difference of Arrival) algorithm on a central coordination server. For equipment The preset position coordinates are obtained through device position measurements during initialization and stored in the topology model database. For equipment Distance from the sound source; The device location coordinates are determined by measurement during system initialization and stored in the static spatial topology model, forming a device location coordinate database.

[0036] like Figure 2 The diagram shown is a structural schematic of a remote control system for a home smart speaker provided in an embodiment of this application. The remote control system for a home smart speaker provided in an embodiment of this application includes: a model building module, a tag acquisition module, a first-level arbitration execution module, a second-level arbitration execution module, a rollback module, and an instruction execution module. The model building module is used to pre-configure and store the binding relationship between each smart speaker device and a logical location identifier, and to build a static spatial topology model. The tag acquisition module is used to enable each smart speaker device to collect its own voice signal in parallel when the user issues a voice command, and to simultaneously activate the gaze detection sensor to acquire its own binary gazed state tag. The first-level arbitration execution module is used by the central coordination server to gather the gaze status tags of all smart speaker devices and execute the first-level arbitration with gaze priority: if it is determined that there is a unique gazed device, it is directly determined as the final response unit; The second-level arbitration execution module is used to initiate the second-level arbitration if the first-level arbitration does not produce a final response unit: when it is determined that there are multiple smart speaker devices being monitored, sound source localization arbitration is only performed between the monitored devices; The rollback module is used to roll back to the position command parsing based on the static spatial topology model to determine the final response unit when it is determined that there is no smart speaker device being watched. The instruction execution module is used to execute voice instructions by the final response unit and broadcast status synchronization messages to other devices to update the status record tables of the other devices and prevent them from responding to conflicts.

[0037] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0038] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0039] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0040] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0041] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0042] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A remote control method for a home smart speaker, characterized in that, Includes the following steps: Pre-configure and store the binding relationships between each smart speaker device and logical location identifier to construct a static spatial topology model; When a user issues a voice command, each smart speaker device collects its own voice signal in parallel and simultaneously acquires its own binary gaze status marker. It aggregates the gaze status tags of all smart speaker devices and performs the first-level arbitration based on gaze priority: if it determines that there is a unique gazed device, it is directly identified as the final response unit; If the first-level arbitration does not produce a final response unit, the second-level arbitration is initiated: when it is determined that there are multiple smart speaker devices being monitored, sound source localization arbitration is only performed between the monitored devices. If it is determined that there is no smart speaker device being observed, the process reverts to the position command parsing based on the static spatial topology model to determine the final response unit. The final response unit executes the voice command and broadcasts a status synchronization message to other devices.

2. The remote control method for a home smart speaker as described in claim 1, characterized in that, The steps to construct a static spatial topology model include: Collect three-dimensional spatial layout data of the home environment and generate a digital map using a laser scanner; Assign a unique logical location identifier to each smart speaker device, and bind the physical coordinates of each smart speaker device to the logical identifier; Based on the relative distance and orientation between each smart speaker device, a topological graph is established, where nodes represent devices and edges represent spatial connectivity and distance weights. Store the binding relationship and topology diagram between the physical coordinates and logical identifiers of each smart speaker device in the database.

3. The remote control method for a home smart speaker as described in claim 1, characterized in that, The steps for synchronously acquiring their respective binary observed state tags include: Get in the time window Internally acquired sensor signals ; Calculate each smart speaker device Weighted integral perception intensity and the preset gaze threshold The binary observed state marker is generated through comparison. ; Right now: ； ； In the formula, For device indexing, Indicates being watched. Indicates that you are not being watched. For equipment The weighted integral perceived intensity, To detect the start time, For the detection time window, For equipment The total number of sensing elements, For the first Each sensing element in Instantaneous signal reading at time 10:

00. For the first The environmental baseline signal of each sensing element For the first Preset weighting coefficients for each sensing element This is the calibrated gaze judgment threshold.

4. The remote control method for a home smart speaker as described in claim 1, characterized in that, The steps for arbitrating sound source localization among monitored devices include: Define the set of multiple observed smart speaker devices as the first candidate set. ; First candidate set Each smart speaker device Acquire the collected voice signal Calculate its location at all possible sound source locations. The steering response power is obtained by applying a phase transformation for weighting. : ； Each smart speaker device Calculate its maximum response power: ； Based on the first candidate set In this study, the smart speaker with the highest response power was selected as the final response device. : ； In the formula, For equipment The The signal from each microphone in the frequency domain Indicates complex conjugation. Angular frequency, Let be the sound source location vector in three-dimensional space. For the sound source in Position to number The propagation delay of each microphone, The imaginary unit, It is a natural constant.

5. The remote control method for a home smart speaker as described in claim 1, characterized in that, The steps for reverting to position command parsing based on the static spatial topology model include: Perform natural language processing on voice commands to extract location keywords; The extracted location keywords are matched with logical location identifiers in the static spatial topology model, and a similarity score is calculated. If a unique smart speaker device is matched, it will be selected as the final responding smart speaker device. If multiple smart speaker devices are matched, the system will further combine sound source localization-assisted arbitration to select the device closest to the sound source as the final response device. If no specific location keyword is specified, the default main device in the topology model will be selected.

6. The remote control method for a home smart speaker as described in claim 5, characterized in that, The steps for calculating the similarity score include: The extracted location keywords are stemmed and normalized to generate standardized keyword vectors; Standardized keywords and logical position identifiers are converted into high-dimensional vector representations using a pre-trained word embedding model; Calculate the cosine similarity between the keyword vector and each logical position identifier vector, and use this as the similarity score. : ； In the formula, This is a vector representation of the keywords. A vector representation of a logical position identifier; If the similarity score is not less than the preset similarity threshold, the corresponding smart speaker device will be included in the candidate list. The device corresponding to the logical location name with the highest similarity score is selected as the candidate response smart speaker device. If multiple smart speaker devices have the same score and it is not less than the similarity threshold, then sound source localization arbitration is initiated among these devices.

7. The remote control method for a home smart speaker as described in claim 5, characterized in that, The steps for selecting the device closest to the sound source as the final response device, combined with sound source localization-assisted arbitration, include: Each smart speaker device collects voice signals and calculates the three-dimensional coordinates of the sound source in space based on a time difference of arrival (TDOA) localization algorithm. ; Based on the device location coordinates stored in the static spatial topology model Calculate the Euclidean distance between each device and the sound source: ； Choose the device closest to the sound source as the final response device, that is: ； In the formula, For the estimated sound source location coordinates, For equipment The preset position coordinates For equipment The distance between the sound source and the sound source.

8. A remote control system for a home smart speaker, used to implement the remote control method for a home smart speaker as described in any one of claims 1-7, characterized in that, include: The module includes a model building module, a tag acquisition module, a first-level arbitration execution module, a second-level arbitration execution module, a rollback module, and an instruction execution module. The model building module is used to pre-configure and store the binding relationship between each smart speaker device and a logical location identifier, and to build a static spatial topology model. The tag acquisition module is used to allow each smart speaker device to collect its own voice signal in parallel and simultaneously acquire its own binary gazed state tag when the user issues a voice command. The first-level arbitration execution module is used to aggregate the gaze status markers of all smart speaker devices and execute the first-level arbitration with gaze priority: if it is determined that there is a unique gazed device, it is directly identified as the final response unit; The second-level arbitration execution module is used to initiate the second-level arbitration if the first-level arbitration does not produce a final response unit: when it is determined that there are multiple smart speaker devices being monitored, sound source localization arbitration is only performed between the monitored devices; The rollback module is used to roll back to the position command parsing based on the static spatial topology model to determine the final response unit when it is determined that there is no smart speaker device being watched. The instruction execution module is used to execute voice instructions by the final response unit and broadcast status synchronization messages to other devices.

Images