Method and system for accurately identifying control instruction of intelligent earphone

Abstract

The invention discloses a method and a system for accurately identifying a control instruction of an intelligent earphone. The method comprises the following steps: establishing low-power-consumption Bluetooth connection between the intelligent earphone and a mobile host end, and performing clock synchronization; iMU data are collected and fused to obtain quaternion attitude data, and a timestamp based on a synchronous clock is added; aligning the attitude data to generate a one-to-one corresponding mapping data list; calculating a relative motion quaternion of the posture data of the intelligent earphone and the mobile host end to form a relative motion data list; and finally, recognizing the earphone control instruction according to the list. By calculating the relative motion of the head relative to the body, the technical problem of head control instruction misrecognition caused by body motion in the prior art is solved, and the accuracy and reliability of posture control of the intelligent earphone in a human body non-static scene are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of smart wearable devices and human-computer interaction technology, and in particular to a method and system for accurately recognizing control commands of smart headphones. Background Technology

[0002] With the rapid development of smart wearable devices, Bluetooth headsets have evolved from simple audio playback devices into multifunctional platforms integrating communication, entertainment, health monitoring, and intelligent interaction. Among these, head motion tracking-based human-computer interaction technology has become a research hotspot in the industry in recent years due to its intuitive and convenient characteristics. Users can answer / hang up calls, switch songs, activate voice assistants, and switch noise cancellation modes by nodding, shaking, or turning their heads, greatly enriching the interactive experience and demonstrating significant value in the field of accessibility assistance. Currently, mainstream head motion tracking solutions rely on the Inertial Measurement Unit (IMU) integrated inside the Bluetooth headset, which typically includes a gyroscope, accelerometer, and magnetometer. The system processes the raw data from the IMU through sensor fusion algorithms (such as Kalman filtering) to calculate the head's attitude angles (such as yaw, pitch, and roll angles) or specific movement patterns, and then maps them into corresponding control commands.

[0003] However, existing technologies cannot effectively distinguish between interference from target head movements and unintentional body movements, resulting in a persistently high misrecognition rate of control commands. The root cause of this problem lies in: First, in daily activities, a user's head and torso move as a coherent whole. For example, when a user walks, runs, bends over, turns around, or experiences bumps while traveling, the movement of their torso (body) inevitably involves and triggers passive head movements. Existing technologies rely solely on the earphone's local IMU sensor, which collects motion data that is actually the absolute movement of the head relative to inertial space—a mixed signal coupled with "active head movements" and "passive body movements." The system cannot extract the user's true intention—the active head movements used for interaction—from this mixed signal, thus misinterpreting a large number of head movements caused by body movement as control commands. Second, from a technical implementation perspective, existing solutions typically set thresholds for the amplitude, speed, or duration of movements at the algorithm level to attempt to avoid false triggers, but this method is largely ineffective. Because intentional head-shaking / nodding movements and unintentional movements caused by body motion (such as the up-and-down head swaying during jogging) may have significant overlap in kinematic features (such as frequency and amplitude), simple threshold-based discrimination algorithms face a dilemma: "increasing the threshold leads to a decrease in command recognition rate, while decreasing the threshold leads to an increase in false trigger rate." Third, existing technologies lack clear guidance on the collaborative processing of motion data from head sensors and body sensors, such as data synchronization, preprocessing, and fusion formulas.

[0004] In conclusion, there is an urgent need for a new and more intelligent method for data acquisition, preprocessing, and corresponding data calculation, which can accurately extract head motion data with clear user intent from complex mixed motion signals, fundamentally solve the problem of misidentification, and improve the accuracy and reliability of interaction.

[0005] The information disclosed in the background section of this invention is intended only to enhance the understanding of the general background of this invention, and should not be construed as an admission or in any way implying that such information constitutes prior art known to those skilled in the art. Summary of the Invention

[0006] This invention proposes a method and system for accurately identifying control commands of smart headphones. By calculating the relative motion of the head to the body, it solves the technical problem of misidentification of head control commands caused by body movement in the prior art, and significantly improves the accuracy and reliability of posture control of smart headphones in non-static human body scenarios.

[0007] In a first aspect, the present invention provides a method for accurately identifying control commands of smart headphones, the method comprising the steps of: The S100 establishes a low-power Bluetooth connection with the mobile host by determining the periodic connection interval through the connection protocol; during each connection cycle, the smart earphone adjusts the clock offset to align with the master clock of the mobile host. S110: The smart earphone MCU reads the raw data from the IMU sensor, fuses it to obtain the earphone quaternion attitude data, adds a timestamp based on the alignment clock offset determined in the most recent connection period, and encapsulates it into an attitude data packet. When the periodic connection interval arrives, the attitude data packet is sent to the mobile host. The mobile host establishes a local cache, reads the raw data from the local IMU sensor, fuses it to obtain the local quaternion attitude data, adds a local timestamp, and puts it into the local cache. S120, the mobile host decapsulates the attitude data packet to obtain the headphone quaternion attitude data, and aligns the headphone quaternion attitude data with the local quaternion attitude data based on the timestamp based on the clock offset and the local timestamp to obtain a one-to-one mapping data list between the two. S130, for each row of data in the mapped data list, calculate the Hamiltonian product of the conjugate data of the headphone quaternion attitude data and the local quaternion attitude data to form a relative motion data list, and identify the headphone control command based on the relative motion data list.

[0008] Preferably, the alignment in step S120 specifically involves: using the timestamp based on the clock offset as the reference time node and marking it on the time axis; using the local timestamp as the time node to be aligned and marking it on the time axis; performing linear interpolation processing on the reference time node based on the local quaternion attitude data to generate the local quaternion attitude data corresponding to the reference time node; and mapping the headphone quaternion attitude data corresponding to the reference time node to the local quaternion attitude data corresponding to the reference time node one-to-one to obtain a mapping data list, thereby completing the alignment operation.

[0009] Preferably, the step S130 of identifying headphone control commands based on the relative motion data list specifically includes: obtaining the average execution time of the action corresponding to the control command; extracting posture feature data based on the user posture data set within the average execution time; labeling the data with command type and feeding it into a recurrent neural network classification model for training to obtain a trained recurrent neural network classification model; feeding the relative motion data list into the trained recurrent neural network classification model to perform model classification prediction; if the prediction confidence is greater than the recognition threshold, using the prediction result as the identification command of the relative motion data list.

[0010] Preferably, before step S130, a first preprocessing is performed on the headphone quaternion attitude data. The first preprocessing includes: acquiring the headphone quaternion attitude data to form a headphone quaternion attitude data list, and determining the periodic variation coefficient CV of the headphone quaternion attitude data list: CV = σ / μ. Where σ is the standard deviation of the values ​​in the headphone quaternion attitude data list, and μ is the average value of the values ​​in the headphone quaternion attitude data list; When the CV is less than the minimum amplitude threshold, the recognition process for this cycle ends.

[0011] Preferably, after obtaining the local quaternion attitude data in step S110 and before putting it into the local cache, a second preprocessing is performed on the local quaternion attitude data sequence. The second preprocessing is as follows: determining the autocorrelation coefficient (ACF) of the local quaternion attitude data sequence, and performing a first stabilization process on the local quaternion attitude data sequence whose ACF exceeds the confidence interval; the second preprocessing also includes: when the autocorrelation coefficient (ACF) does not exceed the confidence interval, determining the motion data change amplitude (dH) based on the local quaternion attitude data sequence, and when the change amplitude (dH) exceeds a preset amplitude threshold (H0), performing a second stabilization process on the data sequence that exceeds the preset amplitude threshold (H0).

[0012] Secondly, embodiments of this disclosure provide a system for accurately identifying control commands for smart headphones, comprising: The connection management module is configured to determine the periodic connection interval through the connection protocol and establish a low-power connection between Bluetooth and the mobile host. During each connection cycle, the clock offset of the smart earphone is adjusted to align with the master clock of the mobile host. A data acquisition module, deployed on the smart earphone, is configured to read raw data from the IMU sensor, fuse it to obtain earphone quaternion attitude data, add a timestamp based on the alignment clock offset determined in the most recent connection period, and encapsulate it into an attitude data packet. When the periodic connection interval arrives, the attitude data packet is sent to the mobile host. A data caching module, deployed on the mobile host, is configured to establish a local cache, read raw data from the local IMU sensor, fuse it to obtain local quaternion attitude data, add a local timestamp, and put it into the local cache. The data alignment module, deployed on the mobile host, is configured to decapsulate the attitude data packet to obtain the headphone quaternion attitude data, and align the headphone quaternion attitude data and the local quaternion attitude data based on the clock offset timestamp and the local timestamp to obtain a one-to-one mapping data list between the two. The instruction recognition module, deployed on the mobile host, is configured to calculate the Hamiltonian product of the conjugate data of the headphone quaternion attitude data and the local quaternion attitude data for each row of data in the mapped data list, forming a relative motion data list, and to recognize headphone control instructions based on the relative motion data list.

[0013] Preferably, the data alignment module is specifically configured to: mark the time axis with the clock offset-based timestamp as the reference time node, mark the time axis with the local timestamp as the time node to be aligned, perform linear interpolation processing on the reference time node based on the local quaternion attitude data to generate the local quaternion attitude data corresponding to the reference time node, and map the headphone quaternion attitude data corresponding to the reference time node to the local quaternion attitude data corresponding to the reference time node one-to-one to obtain a mapping data list, thereby completing the alignment operation.

[0014] Preferably, the instruction recognition module specifically includes: The model training unit is configured to obtain the average execution time of the action corresponding to the control command, extract posture feature data based on the user posture data set within the average execution time, label the command type, and feed it into the recurrent neural network classification model for training to obtain the trained recurrent neural network classification model. The instruction prediction unit is configured to feed the relative motion data list into a trained recurrent neural network classification model to perform model classification prediction. If the prediction confidence is greater than the recognition threshold, the prediction result is used as the recognition instruction for the relative motion data list.

[0015] Preferably, the system further includes a first preprocessing module: the first preprocessing module, deployed on the mobile host, is configured to perform a first preprocessing on the headphone quaternion attitude data before the instruction recognition module runs, including: acquiring the headphone quaternion attitude data to form a headphone quaternion attitude data list, and determining the periodic variation coefficient CV of the headphone quaternion attitude data list. CV = σ / μ, Where σ is the standard deviation of the values ​​in the headphone quaternion attitude data list, and μ is the average value of the values ​​in the headphone quaternion attitude data list; When the CV is less than the minimum amplitude threshold, the recognition process for this cycle ends.

[0016] Preferably, the system further includes a second preprocessing module: the second preprocessing module is deployed on the mobile host and configured to perform a second preprocessing on the local quaternion attitude data sequence after the data caching module obtains the local quaternion attitude data and before it is placed into the local cache, including: determining the autocorrelation coefficient (ACF) of the local quaternion attitude data sequence; performing a first stabilization process on the local quaternion attitude data sequence whose ACF exceeds the confidence interval; and, if the ACF does not exceed the confidence interval, determining the motion data change amplitude dH based on the local quaternion attitude data sequence; and performing a second stabilization process on the data sequence that exceeds the preset amplitude threshold H0 when the change amplitude dH exceeds the preset amplitude threshold H0.

[0017] The solution provided by this invention achieves at least the following technical effects: (1) Based on the core concept of analyzing user intentions by calculating the relative motion of the head to the body, rather than relying on independent absolute motion data, the passive motion interference of the head caused by the user's body movement (such as walking, running, turning, and vehicle bumps) is essentially filtered out, fundamentally and significantly improving the recognition accuracy and reliability of control commands; (2) The entire solution is built on the existing hardware foundation of smart headphones and mobile hosts. Through clock synchronization, data alignment, and various preprocessing methods for data, the accuracy and precision of data processing and command recognition are further guaranteed, providing a superior solution with high efficiency, low power consumption, and high user experience. Users can confidently use head posture for control in mobile scenarios without having to deliberately keep their bodies still, freeing users from the burden of interaction and making head movements a truly natural, convenient, and efficient interaction dimension. The smooth and reliable interaction experience is of great significance for improving the user satisfaction and immersion of smart headphones. Attached Figure Description

[0018] The above and other objects, features and advantages of the present invention will become more apparent from the more detailed description of exemplary embodiments of the invention in conjunction with the accompanying drawings, wherein the same reference numerals generally represent the same parts.

[0019] Figure 1 A flowchart illustrating the steps of a method for accurately identifying smart headphone control commands according to an embodiment of the present invention is shown.

[0020] Figure 2 A block diagram of a system for accurately recognizing control commands for smart headphones according to an embodiment of the present invention is shown. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0022] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a particular order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented, for example, in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “corresponding to,” and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0023] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0024] Example 1: As Figure 1 As shown, this disclosure provides a method for accurately identifying control commands for smart headphones, including: The S100 establishes a low-power Bluetooth connection with the mobile host by determining the periodic connection interval through the connection protocol. During each connection cycle, the smart earphone adjusts the clock offset to align with the master clock of the mobile host.

[0025] The mobile host device, such as a mobile phone or other portable electronic device, is recommended to connect to the Bluetooth headset. Specifically, the Bluetooth module version of the mobile host device should typically be 4.0 or higher to ensure it has a dual-mode mechanism (Bluetooth Dual Mode). This means that a single device can simultaneously support two classic Bluetooth technologies: Bluetooth Classic and Bluetooth Low Energy (BLE). Bluetooth Classic is used for transmitting large, continuous data streams, such as images, files, and music, while Bluetooth Low Energy (BLE) enables extremely low-power, intermittent data transmission, remaining in a sleep state at other times to reduce power consumption. It is primarily used for transmitting smaller amounts of sensor data, such as location data and sensor data.

[0026] In a connected state, the Bluetooth Low Energy (BLE) protocol uses a mechanism called "clock drift" to adjust its clocks to maintain synchronization and ensure accurate communication timing. This underlying characteristic of BLE is not used for communication itself, but rather to establish a unified, high-precision time reference for subsequent sensor data. If the clocks of the two devices are out of sync, their timestamps are incomparable, making it impossible to determine when a movement of the earphone occurred simultaneously with a movement of the phone, rendering all subsequent calculations meaningless. The periodic connection interval determines the frequency of data exchange and synchronization.

[0027] In one embodiment, the BLE protocol stack negotiates a 50ms connection interval. During each connection event, the headset adjusts its clock offset based on data packets sent by the phone, ensuring microsecond-level synchronization with the phone's master clock.

[0028] S110: The smart earphone MCU reads the raw data from the IMU sensor, fuses it to obtain the earphone quaternion attitude data, adds a timestamp based on the alignment clock offset determined in the most recent connection period, and encapsulates it into an attitude data packet. When the periodic connection interval arrives, the attitude data packet is sent to the mobile host. The mobile host establishes a local cache, reads the raw data from the local IMU sensor, fuses it to obtain the local quaternion attitude data, adds a local timestamp, and puts it into the local cache. Step S110 describes the parallel data processing pipeline for the mobile phone and the headset. On the Bluetooth smart headset side, the MCU reads raw data from the IMU (gyroscope, accelerometer) and calculates stable and accurate headset quaternion attitude data using sensor fusion algorithms (such as Kalman filtering and complementary filtering). This data represents the global head orientation, and the timestamp's clock source is based on the clock aligned with the mobile phone in step S100, not the headset's own clock. This is crucial for ensuring timestamp comparability. The data pairs are encapsulated into packets and sent to the mobile host when the next agreed-upon connection event occurs.

[0029] The phone also reads its own IMU data and fuses it to obtain the phone's quaternion pose data. This represents the global orientation of the body (or the platform on which the phone is located). The phone's local quaternion pose data is timestamped with the phone's local clock (due to the synchronization of the S100, this local clock is aligned with the headphone's clock) and stored in a first-in-first-out (FIFO) buffer, waiting to be matched with the headphone data.

[0030] In one implementation scenario, such as subway swaying, vehicle start-up, and braking, the user's body sways (e.g., leaning forward or turning). The phone's IMU senses this motion, fuses it into a posture quaternion (e.g., indicating the phone is leaning forward with the body), and stores it in a cache with a local timestamp. At the same moment, the user nods. The headset's IMU senses the head's pitch relative to the neck, and the MCU fuses it into a posture quaternion representing the head tilting downward. The headset's firmware timestamps this data based on a clock synchronized with the phone and sends the data packet to the phone when the next connection event occurs.

[0031] S120, the mobile host decapsulates the attitude data packet to obtain the headphone quaternion attitude data, and aligns the headphone quaternion attitude data with the local quaternion attitude data based on the timestamp based on the clock offset and the local timestamp to obtain a one-to-one mapping data list between the two. In a preferred embodiment, the alignment in step S120 specifically involves: using the timestamp based on the clock offset as the reference time node and marking it on the time axis; using the local timestamp as the time node to be aligned and marking it on the time axis; performing linear interpolation processing on the reference time node based on the local quaternion attitude data to generate the local quaternion attitude data corresponding to the reference time node; and mapping the headphone quaternion attitude data corresponding to the reference time node to the local quaternion attitude data corresponding to the reference time node one-to-one to obtain a mapping data list, thereby completing the alignment operation.

[0032] For example, if a mobile phone receives a headphone data packet with a timestamp of 1234.126s, the phone searches its local cache and finds two local quaternion attitude data corresponding to the nearest timestamps of 1234.120s and 1234.130s. The phone then uses a linear interpolation algorithm to estimate the phone's attitude data at the precise moment of 1234.126s based on these two data points, and combines this data with the headphone data packet from 1234.126s to form a data pair and put it into a mapping list.

[0033] S130, for each row of data in the mapped data list, calculate the Hamiltonian product of the conjugate data of the headphone quaternion attitude data and the local quaternion attitude data to form a relative motion data list, and identify the headphone control command based on the relative motion data list.

[0034] For each pair of data in the mapping list, calculate their relative motion, the geometric meaning of which is: remove the influence of the phone (body) global motion, and obtain the pure motion that only describes the head's movement relative to the body.

[0035] In a preferred embodiment, prior to step S130, a first preprocessing step is performed on the headphone quaternion attitude data. This first preprocessing includes: acquiring the headphone quaternion attitude data to form a headphone quaternion attitude data list, and determining the periodic variation coefficient (CV) of the headphone quaternion attitude data list. CV = σ / (μ×a×T), Where σ is the standard deviation of the headphone quaternion attitude data list, μ is the average value of the headphone quaternion attitude data list, T is the connection interval period, and a is the weight scaling factor. When the CV is less than the minimum amplitude threshold, the recognition process for this cycle ends.

[0036] The first preprocessing step is used for cases where the head movement amplitude is small. In this case, regardless of the body movement data, command recognition does not need to be considered.

[0037] In a preferred embodiment, after obtaining the local quaternion attitude data in step S110 and before placing it into the local cache, a second preprocessing is performed on the local quaternion attitude data sequence. The second preprocessing is: determining the autocorrelation coefficient (ACF) of the local quaternion attitude data sequence, and performing a first stabilization process on local quaternion attitude data sequences whose ACF exceeds the confidence interval (e.g., 0.95). The first stabilization process specifically includes: replacing the original value with the average value of the data in the quaternion attitude data sequence; or replacing the original value by reducing each data value by a percentage; or discarding the original value by replacing the original value with a null value, etc.

[0038] The second preprocessing described above is for the periodic movement of the mobile host. In this case, for example, when the mobile phone is in a pocket, trouser pocket, backpack, or handbag, it moves periodically back and forth based on walking; or when the mobile phone is held in the hand, it moves periodically back and forth based on running. In this case, there is no need to consider the local quaternion posture data detected by the mobile phone to prevent data interference.

[0039] In a preferred embodiment, the second preprocessing further includes: if the autocorrelation coefficient (ACF) does not exceed the confidence interval (0.95), determining the motion data change amplitude (dH) based on the local quaternion attitude data sequence; if the change amplitude (dH) exceeds a preset amplitude threshold (H0), then performing a second stabilization process on the data sequence that exceeds the preset amplitude threshold (H0); the second stabilization process specifically involves: replacing the original value with a neighboring value of the data value whose motion data change amplitude (dH) exceeds the preset amplitude threshold (H0) in the quaternion attitude data sequence; or discarding the original value by replacing it with a null value, etc.

[0040] The second preprocessing described above is used for periodic motion cases on the mobile host. However, when the motion amplitude is too large to represent the body's motion characteristics, the non-periodic motion cases on the mobile host are also considered. In this case, it is not necessary to consider the local quaternion pose data detected by the mobile phone to prevent data interference.

[0041] In general, the principle of the second preprocessing is to correlate the motion data monitored by the mobile host with the actual motion data of the body and to eliminate periodic motion data interference, so that the motion data is as consistent as possible with the physiological phenomena of body movement, and appears reasonably and randomly. The processed data can significantly increase the accuracy of motion command recognition. Under the above principles, the second preprocessing also includes the possibility of other further improvements.

[0042] The first and second preprocessing processes significantly improved the accuracy of control command recognition.

[0043] In a preferred embodiment, before step S130, the method further includes: obtaining the average execution time of the action corresponding to the control command, extracting posture feature data based on the user posture data set within the average execution time, labeling it with command type tags, and feeding it into a recurrent neural network classification model for training to obtain a trained recurrent neural network classification model; further, the step of identifying the headphone control command based on the relative motion data list specifically involves: feeding the relative motion data list into the trained recurrent neural network classification model to perform model classification prediction, and if the prediction confidence is greater than the recognition threshold, using the prediction result as the identification command of the relative motion data list.

[0044] In a preferred embodiment, the average execution time can be the average execution time of actions corresponding to execution commands from users within a specific range collected by the system, or the average execution time of actions corresponding to execution commands from the current user on the collected device. The user posture data set can be the user posture data set of users within a specific range, or the historical user posture data set of actions corresponding to execution commands from the current user on the collected device.

[0045] In another preferred embodiment, instead of using a recurrent neural network classification model, simple geometric rules are set for different types of actions, limiting the action features through time and space thresholds, including: (1) recognizing the "nodding" command: checking whether the pitch angle first increases positively and then decreases within a short period of time (e.g., 0.5 seconds), and the change exceeds a threshold (e.g., 20 degrees); (2) recognizing the "shaking" command: checking whether the yaw angle changes alternately left and right within a short period of time, and exceeds the threshold to complete one cycle. The setting of geometric rules needs to be dynamically adjusted according to the user's personality to improve the accuracy of command recognition and user experience.

[0046] It accurately identifies the smart control commands of Bluetooth headsets based on the relative movement of the head and body, which is more in line with the user's operating habits and can significantly reduce the misrecognition of commands caused by body movements.

[0047] Example 2: Figure 2 As shown, the present invention also provides a system for accurately recognizing control commands of smart headphones, comprising: A system for accurately recognizing control commands of smart headphones, characterized in that the system comprises: The connection management module is configured to determine the periodic connection interval through the connection protocol and establish a low-power connection between Bluetooth and the mobile host. During each connection cycle, the clock offset of the smart earphone is adjusted to align with the master clock of the mobile host. A data acquisition module, deployed on the smart earphone, is configured to read raw data from the IMU sensor, fuse it to obtain earphone quaternion attitude data, add a timestamp based on the alignment clock offset determined in the most recent connection period, and encapsulate it into an attitude data packet. When the periodic connection interval arrives, the attitude data packet is sent to the mobile host. A data caching module, deployed on the mobile host, is configured to establish a local cache, read raw data from the local IMU sensor, fuse it to obtain local quaternion attitude data, add a local timestamp, and put it into the local cache. The data alignment module, deployed on the mobile host, is configured to decapsulate the attitude data packet to obtain the headphone quaternion attitude data, and align the headphone quaternion attitude data and the local quaternion attitude data based on the clock offset timestamp and the local timestamp to obtain a one-to-one mapping data list between the two. The instruction recognition module, deployed on the mobile host, is configured to calculate the Hamiltonian product of the conjugate data of the headphone quaternion attitude data and the local quaternion attitude data for each row of data in the mapped data list, forming a relative motion data list, and to recognize headphone control instructions based on the relative motion data list.

[0048] In a preferred embodiment, the data alignment module is specifically configured to: mark the time axis with the clock offset-based timestamp as the reference time node, mark the time axis with the local timestamp as the time node to be aligned, perform linear interpolation processing on the reference time node based on the local quaternion attitude data to generate the local quaternion attitude data corresponding to the reference time node, and map the headphone quaternion attitude data corresponding to the reference time node to the local quaternion attitude data corresponding to the reference time node one-to-one to obtain a mapping data list, thereby completing the alignment operation.

[0049] In a preferred embodiment, the instruction recognition module specifically includes: The model training unit is configured to obtain the average execution time of the action corresponding to the control command, extract posture feature data based on the user posture data set within the average execution time, label the command type, and feed it into the recurrent neural network classification model for training to obtain the trained recurrent neural network classification model. The instruction prediction unit is configured to feed the relative motion data list into a trained recurrent neural network classification model to perform model classification prediction. If the prediction confidence is greater than the recognition threshold, the prediction result is used as the recognition instruction for the relative motion data list.

[0050] In a preferred embodiment, the system further includes a first preprocessing module: the first preprocessing module, deployed on the mobile host, is configured to perform a first preprocessing on the headphone quaternion attitude data before the instruction recognition module runs, including: acquiring the headphone quaternion attitude data to form a headphone quaternion attitude data list, and determining the periodic variation coefficient CV of the headphone quaternion attitude data list. CV = σ / μ, Where σ is the standard deviation of the values ​​in the headphone quaternion attitude data list, and μ is the average value of the values ​​in the headphone quaternion attitude data list; When the CV is less than the minimum amplitude threshold, the recognition process for this cycle ends.

[0051] In a preferred embodiment, the system further includes a second preprocessing module: the second preprocessing module is deployed on the mobile host and configured to perform a second preprocessing on the local quaternion attitude data sequence after the data caching module obtains the local quaternion attitude data and before it is placed into the local cache, including: determining the autocorrelation coefficient (ACF) of the local quaternion attitude data sequence; performing a first stabilization process on the local quaternion attitude data sequence whose ACF exceeds the confidence interval; and, if the ACF does not exceed the confidence interval, determining the motion data change amplitude dH based on the local quaternion attitude data sequence; and performing a second stabilization process on the data sequence that exceeds the preset amplitude threshold H0 when the change amplitude dH exceeds the preset amplitude threshold H0.

[0052] The solution provided by this invention achieves at least the following technical effects: (1) Based on the core concept of analyzing user intentions by calculating the relative motion of the head to the body, rather than relying on independent absolute motion data, the passive motion interference of the head caused by the user's body movement (such as walking, running, turning, and vehicle bumps) is essentially filtered out, fundamentally and significantly improving the recognition accuracy and reliability of control commands; (2) The entire solution is built on the existing hardware foundation of smart headphones and mobile hosts. Through clock synchronization, data alignment, and various preprocessing methods for data, the accuracy and precision of data processing and command recognition are further guaranteed, providing a superior solution with high efficiency, low power consumption, and high user experience. Users can confidently use head posture for control in mobile scenarios without having to deliberately keep their bodies still, freeing users from the burden of interaction and making head movements a truly natural, convenient, and efficient interaction dimension. The smooth and reliable interaction experience is of great significance for improving the user satisfaction and immersion of smart headphones.

[0053] According to one embodiment, a program product, such as a machine-readable medium, is provided. The machine-readable medium may have instructions (i.e., the elements implemented in software as described above), which, when executed by a machine, cause the machine to perform the above-described combinations of the various embodiments of this specification. Figure 1 The various operations and functions described. Specifically, a system or apparatus equipped with a readable storage medium storing software program code that implements the functions of any of the embodiments described above, and enabling the computer or processor of the system or apparatus to read and execute the instructions stored in the readable storage medium.

[0054] In this case, the program code read from the readable medium itself can perform the functions of any of the above embodiments, and therefore the machine-readable code and the readable storage medium storing the machine-readable code constitute a part of this specification.

[0055] Examples of readable storage media include floppy disks, hard disks, magneto-optical disks, optical disks (such as CD-ROM, CD-R, CD-RW, DVD-ROM, DVD-RAM, DVD-RW, DVD-RW), magnetic tapes, non-volatile memory cards, and ROMs. Alternatively, program code can be downloaded from a server computer or the cloud via a communication network.

[0056] Those skilled in the art will understand that the various embodiments disclosed above can be modified and varied without departing from the spirit of the invention. Therefore, the scope of protection of this specification should be defined by the appended claims.

[0057] It should be noted that not all steps and units in the above process and system structure diagrams are mandatory; some steps or units can be omitted according to actual needs. The execution order of each step is not fixed and can be determined as needed. The device structure described in the above embodiments can be a physical structure or a logical structure. That is, some units may be implemented by the same physical client, or some units may be implemented by multiple physical clients, or they may be jointly implemented by certain components in multiple independent devices.

[0058] In the above embodiments, the hardware units or modules can be implemented mechanically or electrically. For example, a hardware unit, module, or processor may include permanent dedicated circuitry or logic (such as a dedicated processor, FPGA, or ASIC) to perform the corresponding operation. The hardware unit or processor may also include programmable logic or circuitry (such as a general-purpose processor or other programmable processor), which can be temporarily configured by software to perform the corresponding operation. The specific implementation method (mechanical, dedicated permanent circuitry, or temporarily configured circuitry) can be determined based on cost and time considerations.

[0059] The specific embodiments described above with reference to the accompanying drawings are exemplary embodiments, but do not represent all embodiments that can be implemented or fall within the scope of the claims. The term "exemplary" as used throughout this specification means "serving as an example, instance, or illustration" and does not imply that it is "preferred" or "advantageous" compared to other embodiments. Specific details are included to provide an understanding of the described techniques. However, these techniques can be practiced without these specific details. In some instances, well-known structures and apparatuses are shown in block diagram form to avoid obscuring the concepts of the described embodiments.

[0060] The foregoing description of this disclosure is provided to enable any person skilled in the art to implement or use this disclosure. Various modifications to this disclosure will be apparent to those skilled in the art, and the general principles applicable herein can be applied to other variations without departing from the scope of this disclosure. Therefore, this disclosure is not limited to the examples and designs described herein, but is consistent with the widest scope of the principles and novel features disclosed herein.

Claims

1. A method for accurately identifying control commands of smart headphones, the method comprising the steps of: The S100 establishes a low-power Bluetooth connection with the mobile host by determining the periodic connection interval through the connection protocol; during each connection cycle, the smart earphone adjusts the clock offset to align with the master clock of the mobile host. S110: The smart earphone MCU reads the raw data from the IMU sensor, fuses it to obtain the earphone quaternion attitude data, adds a timestamp based on the alignment clock offset determined in the most recent connection period, and encapsulates it into an attitude data packet. When the periodic connection interval arrives, the attitude data packet is sent to the mobile host. The mobile host establishes a local cache, reads the raw data from the local IMU sensor, fuses it to obtain the local quaternion attitude data, adds a local timestamp, and puts it into the local cache. S120, the mobile host decapsulates the attitude data packet to obtain the headphone quaternion attitude data, and aligns the headphone quaternion attitude data with the local quaternion attitude data based on the timestamp based on the clock offset and the local timestamp to obtain a one-to-one mapping data list between the two. S130, for each row of data in the mapped data list, calculate the Hamiltonian product of the conjugate data of the headphone quaternion attitude data and the local quaternion attitude data to form a relative motion data list, and identify the headphone control command based on the relative motion data list.

2. The method as described in claim 1, characterized in that: The alignment in step S120 specifically involves: using the timestamp based on the clock offset as the reference time node and marking it on the time axis; using the local timestamp as the time node to be aligned and marking it on the time axis; performing linear interpolation processing on the reference time node based on the local quaternion attitude data to generate the local quaternion attitude data corresponding to the reference time node; and mapping the headphone quaternion attitude data corresponding to the reference time node to the local quaternion attitude data corresponding to the reference time node one-to-one to obtain a mapping data list, thereby completing the alignment operation.

3. The method as described in claim 1, characterized in that, The step S130 of identifying headphone control commands based on the relative motion data list specifically includes: obtaining the average execution time of the action corresponding to the control command; extracting posture feature data based on the user posture data set within the average execution time; labeling the data with command type and feeding it into a recurrent neural network classification model for training to obtain a trained recurrent neural network classification model; feeding the relative motion data list into the trained recurrent neural network classification model to perform model classification prediction; if the prediction confidence is greater than the recognition threshold, using the prediction result as the recognition command of the relative motion data list.

4. The method as described in claim 1, characterized in that, Before step S130, a first preprocessing is performed on the headphone quaternion attitude data. The first preprocessing includes: acquiring the headphone quaternion attitude data to form a headphone quaternion attitude data list, and determining the periodic variation coefficient CV of the headphone quaternion attitude data list: CV = σ / μ. Where σ is the standard deviation of the values ​​in the headphone quaternion attitude data list, and μ is the average value of the values ​​in the headphone quaternion attitude data list; When the CV is less than the minimum amplitude threshold, the recognition process for this cycle ends.

5. The method as described in claim 1, characterized in that, After obtaining the local quaternion attitude data in step S110 and before putting it into the local cache, a second preprocessing is performed on the local quaternion attitude data sequence. The second preprocessing is as follows: the autocorrelation coefficient (ACF) of the local quaternion attitude data sequence is determined, and a first stabilization process is performed on the local quaternion attitude data sequence whose ACF exceeds the confidence interval. The second preprocessing also includes: if the autocorrelation coefficient (ACF) does not exceed the confidence interval, the amplitude of motion data change (dH) is determined based on the local quaternion attitude data sequence. If the amplitude of change (dH) exceeds a preset amplitude threshold (H0), a second stabilization process is performed on the data sequence that exceeds the preset amplitude threshold (H0).

6. A system for accurately recognizing control commands of smart headphones, characterized in that, The system includes: The connection management module is configured to determine the periodic connection interval through the connection protocol and establish a low-power connection between Bluetooth and the mobile host. During each connection cycle, the clock offset of the smart earphone is adjusted to align with the master clock of the mobile host. A data acquisition module, deployed on the smart earphone, is configured to read raw data from the IMU sensor, fuse it to obtain earphone quaternion attitude data, add a timestamp based on the alignment clock offset determined in the most recent connection period, and encapsulate it into an attitude data packet. When the periodic connection interval arrives, the attitude data packet is sent to the mobile host. A data caching module, deployed on the mobile host, is configured to establish a local cache, read raw data from the local IMU sensor, fuse it to obtain local quaternion attitude data, add a local timestamp, and put it into the local cache. The data alignment module, deployed on the mobile host, is configured to decapsulate the attitude data packet to obtain the headphone quaternion attitude data, and align the headphone quaternion attitude data and the local quaternion attitude data based on the clock offset timestamp and the local timestamp to obtain a one-to-one mapping data list between the two. The instruction recognition module, deployed on the mobile host, is configured to calculate the Hamiltonian product of the conjugate data of the headphone quaternion attitude data and the local quaternion attitude data for each row of data in the mapped data list, forming a relative motion data list, and to recognize headphone control instructions based on the relative motion data list.

7. The system as described in claim 6, characterized in that, The data alignment module is specifically configured to: The timeline is marked with the clock offset-based timestamp as the reference time node and the local timestamp as the time node to be aligned. Linear interpolation is performed on the reference time node based on the local quaternion attitude data to generate the local quaternion attitude data corresponding to the reference time node. The headphone quaternion attitude data corresponding to the reference time node is mapped one-to-one with the local quaternion attitude data corresponding to the reference time node to obtain a mapping data list, thereby completing the alignment operation.

8. The system as described in claim 6, characterized in that, The instruction recognition module specifically includes: The model training unit is configured to obtain the average execution time of the action corresponding to the control command, extract posture feature data based on the user posture data set within the average execution time, label the command type, and feed it into the recurrent neural network classification model for training to obtain the trained recurrent neural network classification model. The instruction prediction unit is configured to feed the relative motion data list into a trained recurrent neural network classification model to perform model classification prediction. If the prediction confidence is greater than the recognition threshold, the prediction result is used as the recognition instruction for the relative motion data list.

9. The system as described in claim 6, characterized in that, The system further includes a first preprocessing module: the first preprocessing module, deployed on the mobile host, is configured to perform a first preprocessing on the headphone quaternion attitude data before the instruction recognition module runs, including: acquiring the headphone quaternion attitude data to form a headphone quaternion attitude data list, and determining the periodic variation coefficient CV of the headphone quaternion attitude data list. CV = σ / μ, Where σ is the standard deviation of the values ​​in the headphone quaternion attitude data list, and μ is the average value of the values ​​in the headphone quaternion attitude data list; When the CV is less than the minimum amplitude threshold, the recognition process for this cycle ends.

10. The system as described in claim 6, characterized in that, The system further includes a second preprocessing module: the second preprocessing module is deployed on the mobile host and configured to perform a second preprocessing on the local quaternion attitude data sequence after the data caching module obtains the local quaternion attitude data and before it is placed into the local cache, including: determining the autocorrelation coefficient (ACF) of the local quaternion attitude data sequence; performing a first stabilization process on the local quaternion attitude data sequence whose ACF exceeds the confidence interval; and, if the ACF does not exceed the confidence interval, determining the motion data change amplitude dH based on the local quaternion attitude data sequence; and performing a second stabilization process on the data sequence that exceeds the preset amplitude threshold H0 when the change amplitude dH exceeds the preset amplitude threshold H0.

Images