Description:
The following specification particularly describes the nature of the invention and how it is performed:

FIELD OF THE INVENTION
[001] The present invention relates generally to the machine learning and artificial intelligence (AI) field of computer science and more precisely to the area pertains to the development of an innovative and reliable system for monitoring and correcting racket sports activities. This system employs an innovatively designed sensor module seamlessly integrated within the racket's grip, facilitating real-time data acquisition during gameplay. Additionally, the invention more particularly relates to interfaces with a smartphone application to provide comprehensive feedback and correction recommendations based on the data from the sensor module in the grip.
BACKGROUND OF THE INVENTION
[002] The following description facilitates insights that may be useful for comprehending the present invention. Notably, this disclosure does not provide an acknowledgement that any of the details provided herein constitute prior art or are pertinent to the presently asserted invention. Furthermore, no publication mentioned, either explicitly or implicitly, is considered prior art about the invention being claimed.
[003] Further, the systems discussed in this section are potential directions that could be explored, without being inherently indicative of prior conception or pursuit. Henceforth, unless indicated otherwise, it should not be assumed that any of the approaches outlined in this section automatically qualify as prior art solely by their inclusion here.

[004] Racket-based gaming activities, such as badminton, table tennis, and lawn tennis, have become popular as sports activities and the simplest mechanism of staying fit. Thus, it attracts a large number of novice population as well as well-trained sports persons. Moreover, any sports activities, including racket-based, require continuous monitoring and supervision to achieve the desired goal of precise (or winning) performance Traditional coaching methods often rely on subjective observations, which may not capture subtle nuances in a player's technique. In addition, it also holds economical and time-specific training constraints. It motivates a desire to leverage advanced technology to enhance performance and training in racket sports. By integrating sensor technology directly into the grip of the racket, we aim to obtain precise, real-time data on key metrics like grip strength, angle, and pressure.
[005] Sensory data analytics, an advanced analytical approach leveraging information gathered from sensors, along with the application of machine learning and statistical algorithms, has found extensive use across various domains for quite some time. In sports and activities of daily living, sensory data analytics has been particularly useful for extracting information about the categories of the activities (or shorts), the lag time from true behaviour, and creating personalized profiles of each individual. Therefore, helps the individual to analyse their progress and deviations with effective feedback.
[006] Profiling, a method involving the observation and analysis of movements such as those of the hands and legs, as well as behavioural patterns of individuals over some time, is a widely used approach in sensory data analytics. Traditional profiling techniques focus primarily on the data collected from only a few individuals from different regions of specific weather and environmental conditions. For example, the data collected from some individuals in the USA is used to profile the players in India. The collected data may be from different age groups and may be stale as collected off-situ long back. The technique compares the current profile of the users with the profiled data of different individuals from diverse locations to perform analysis and suggest corrections.

[007] Nevertheless, the traditional methods for analysing sensory data and creating profiles have certain limitations in terms of their range and utility. Despite their effectiveness in performing predictive analytics on previously collected sensory data, they have generally focused on off-situ or data collected in diverse locations and populations, overlooking an important aspect of real-time and in-situ data collection, while preserving the personalized features of the players.

[008] Traditional data analysis techniques have not extended the use of predictive analytics on handheld devices (Smartphones). The omission represents a gap in in-situ data analysis approaches, leaving room for improvement in the field. Extending predictive analytics to smartphones could further enhance the accuracy and efficiency of the racket activities, offering a more holistic view of players' performance and providing one-touch correction/ improvement feedback.

[009] Accordingly, based on aforesaid facts, there remains a requirement apart from the prior art for intelligently sensor data collection and in-situ analysis of racket-based activities and techniques thereof. Therefore, it would be useful and desirable to have an end-to-end indigenous system, method, apparatus and interfaces to meet the above-mentioned needs.

[010] Various methodologies and devices have been innovated to facilitate the monitoring of sensor-based activities, with one important approach being the utilization of inertial sensors for gait activity recognition. This specialized field involves the analysis of sensory data acquired from inertial sensors to recognize and identify patterns associated with an individual's gait. Inertial sensors, such as accelerometers, magnetometers, and gyroscopes, capture crucial details of movements, enabling the extraction of valuable information related to the characteristics and dynamics of the gait. Similar mechanisms are discussed in the prior art for sensors-based human activity recognition.
US11478167B2 introduces innovative techniques, systems, and sensors for activity recognition, recording, analysis, and control. The patent uses multiple sensory tags with unique identification and data transfer attributes, strategically placed on the user's body, clothing, personal effects, exercise equipment, and other relevant locations. These tags generate positional, movement, orientation, and acceleration data, which is transmitted to a control system. The system defines a personal activity space, selectively samples data, and employs a simplified form of object recognition to identify, record, and analyze user activities.
CN110998696B introduces a data-driven mobile skill training system that employs mobile skill assessment and diagnosis across various levels of the human mobile system hierarchy. The system tailors training goals based on the assessment, offering synthesized enhancements to aid users in achieving their training objectives. Additionally, the system incorporates features for tracking and managing the learning process, providing a comprehensive approach to mobile skill development.

US10441212B2 describes a method for determining the position and state of a personal activity monitor worn by a user. The activity monitor includes a motion sensor (3-axis accelerometer) and a processor unit. The method involves evaluating the changes in the predominant axis over time, and comparing it against a predetermined threshold to detect wrist-worn conditions. Additionally, the method considers the current activity level and temperature criteria to infer whether the user is wearing the device, entered a sleep phase, or emerged from sleep.

US20210142881A1 pertains to athletic performance sensing and tracking systems designed to measure and sense athletic performance data while providing users with options for creating diverse workouts and accessing relevant information. These systems facilitate user engagement by allowing customization of workouts, selection of media content during athletic activities, and maintaining user motivation. The incorporation of user feedback, combined with objective workout data, enables the control of various workout routine features, selection of music or media content, and adjustment of future workout routines and presented media content.
US20180049670A1 develops a wireless in-shoe physical activity monitoring apparatus, featuring left and right shoe sensor systems. Each shoe sensor system comprises pressure sensing elements, an accelerometer, and a control circuit equipped with essential components such as a power source circuit, clock circuit, processing module, memory, wireless communication transceiver, sensor communication links, and an accelerometer communication link. The apparatus functions by sampling data from the pressure sensing element to generate foot force data and acquiring three-dimensional foot data from the accelerometer. Through wireless communication transceivers, the apparatus transmits outbound radio frequency (RF) signals containing information on both foot force data and three-dimensional foot data, enabling comprehensive monitoring of physical activity.
US11679300B2 discloses systems, media, and methods designed for quantifying and monitoring exercise and motion parameters in real-time, involving data acquisition, analysis, and the provision of scientifically valid, clinically relevant, and actionable diagnostic feedback. The embodiments focus on delivering baseline-adjusted real-time feedback to users, determining activity types, receiving data from motion sensors indicating a time-dependent series of three-axis acceleration and orientation data and presenting a graphical user interface with a real-time representation of the received data. The real-time representation includes a scaled view of at least one dimension of the data based on updated baseline adjustments, enhancing the user's understanding of their exercise or motion performance.
US20190388728A1 describes systems and methods that utilize a wearable sensor for sports action recognition and assessment. The wearable sensor employs a motion sensor data-driven framework, providing real-time kinematical analysis during active competition or training sessions. The system is designed to recognize specific sports actions performed by athletes and assess their skill levels based on the analysis of these actions. The motion sensor data processing platform utilizes a micro inertial measurement unit configuration within the wearable sensor to capture and report data related to the movement of an athlete's limb, facilitating comprehensive analysis by a processor-based system.
US9901776B2 discloses a system for a racket that integrates an inertial sensor, a processor, and a memory device. The inertial sensor, featuring both accelerometer and gyro arrays with three degrees of freedom each, is attached to the racket. The processor interprets signals from these arrays to create stroke profiles detailing the rocket's acceleration and rotation. These stroke profiles are then stored in the memory device, facilitating comprehensive motion tracking and analysis for racket sports.
US20200381021A1 pertains to a method designed to enhance user performance in a sporting activity. The method involves the identification and display of video data of a user, either independently or synchronized with other data, such as concurrently collected biometric data. Key features include the separate storage of biometric data for efficient searching, enabling the rapid identification of values associated with biometric events. Utilizing biometric time stamps linked to identified values facilitates the identification of corresponding video time-frames. These time-frames are then employed to play the video data, either alone or in conjunction with biometric data, starting at specific times, such as when the event occurred or shortly before. The system offers a comprehensive approach to analyzing and improving user performance by integrating video and biometric data seamlessly.
Although several sensor-based activity recognitions are available in the prior art, most of them used sensory data from multiple sensors and performed respective analyses. Some of them focus on activity recognition, and very few use feedback. Next, the work requires wearables sensors or sensors mounted on different devices. Hence, a system for racket activity recognition and correction is required that uses embedded sensors in the racket itself and leverages smartphones for in-situ data collection, processing, and analysis results. In addition, it provides corrective measures on the go using obtained feedback from the data analysis without any human intervention.
SUMMARY OF THE PRESENT INVENTION
[011] Given the foregoing disadvantages inherent in the known types of conventional sensors-based activities recognition, methods and techniques, which are now present in the prior art, the present invention provides an embedded inertial sensors-based approach for in-situ data collection, analysis, and correction. It has all the advantages of the prior art and none of the disadvantages.

[011] The present invention discloses an embedded sensor probe design to be inserted into the racket without interfering with the regular playing comfort of the players as depicted in Fig. 1. Afterwards, it involves predictive analysis of the in-situ data collected from the embedded sensor probe and smartphone. The block diagram of the prototype system is illustrated in Fig. 2 which consists of an IMU sensor, Li-on battery, microcontroller with Wi-Fi/Bluetooth, and a changing module. Furthermore, the deep learning model for prediction on the smartphone is compressed and its performance is enhanced with advanced techniques like knowledge distillation. Moreover, the invention not only monitors the activities but automatically suggests corrective measures to the individual players. This invention involves the following important steps from data collection to suggested correction:

[012] Monitoring: The system keeps track of player-sorting hitting patterns using the embedded sensor probed in the racket while playing different racket-based sports. These sports could be badminton, table tennis, lean tennis or any other racket-based sports activity.
[013] Generating a Predictive Model: Utilizing data on racket activities, the system develops a predictive model by amalgamating the short-hitting pattern profiles of players with those of the coach/expert pre-existing in the database. This process results in the creation of an analytical representation of stroke patterns, enabling the detection and correction of errors within the sport.

[014] Compressing and Storing the Predictive Model: To create a high-performing predictive model that works well on smartphones, we train a predictive model (deep neural networks) with collected and labelled data. However, to make sure it fits on smartphones with limited resources, we need to reduce the model's size. This is where filter pruning and weight factorization come in – they help shrink the model size, allowing it to be easily used on smartphones for analysis, prediction, and suggesting corrective actions, consuming limited resources. However, reducing the size can affect performance, but this invention addresses this by using knowledge distillation from a larger model trained before compression.
[015] Data Transmission: The smartphone uses a communication network, like Bluetooth, to get data from the sensor in the racket. It sends the vibration signal to the phone when there is a greater deviation from the right position.

[016] Generating Predictive Information: Based on the received data, the predictive model deployed on the smartphone generates predictive information about the short played and its deviation angle from the true position ( i.e., position of expert/trainer). This is achieved by evaluating
the input data against the available trained data utilized by the model during training.

[017] This novel approach to in-situ captivity monitoring improved accuracy and efficiency, providing a robust tool for sports players to achieve trainer/coach-free training. The predictive model, combined with real-time data analysis, allows the system to effectively detect correct short plays of the user and flag potential deviation, providing both cost-effectiveness and enhanced efficiency of the players.

[018] Before delving into the detailed explanation of at least one aspect of the invention, it's important to clarify that the application of the invention is not confined to the specific rules and arrangements presented in the following description or depicted in the drawings. The invention can serve other purposes and can be implemented in different ways, depending on the player's requirements. Additionally, the language and terminology used here are for descriptive purposes and should not be considered restrictive.
[019] The invention, including its various components, distinctive features, and advantages, is described in detail in this disclosure. To gain a better understanding of how the invention operates and the specific goals it achieves, please refer to the accompanying drawings and descriptive content, which illustrate preferred embodiments of the invention.
The collective components, along with additional features inherent to this innovation, are explicitly detailed in the disclosure. For a deeper understanding of the innovation, its operational advantages, and the specific objectives attained through its implementation, it is advisable to consult the accompanying illustrations and explanatory materials. These resources showcase preferred embodiments, providing a comprehensive insight into the innovation's distinctive attributes.
BRIEF DESCRIPTION OF THE DRAWINGS
[020] After carefully going through the detailed explanation of the innovation, it will be easier to understand and discover additional goals not mentioned before. This explanation relates to the pictures in the annexe, wherein:
[021] FIG. 1, illustrates a block diagram showing various entities for a racket activity monitoring system and the method thereof, following an embodiment of the present invention. It also contains different short positions 101 of the racket-based activities.

[022] FIG. 2, illustrates the block diagram of the sensor probe to be embedded in the rod of the racket with IMU sensor 203, Li-on battery 201, micro-controller with Wi-Fi/Bluetooth 204, and changing module 202, following an embodiment of the present invention.
[023] FIG. 3, represents the acceleration responses in three axes 301 302 303 of trained/active (dark line) and untrained/average (shadow line) players.
[024] FIG. 4, represents the gyroscope responses in three axes 401 402 403 of trained/active (dark line) and untrained/average (shadow line) players.
[025] FIG. 5, illustrates the semantic diagram for the entire training (Phase I) and testing (Phase II) involved in the proposed racket-based activity recognition system.
DETAILED DESCRIPTION OF THE INVENTION
[026] The following sections of this article will provide various embodiments
of the current invention with references to the accompanying drawings. The reference numbers in the illustrations correspond to similar elements throughout the descriptions. It's important to note that the invention is not limited to the described embodiment; there are several other possible embodiments. The included embodiment aims to ensure a comprehensive and complete disclosure, providing individuals with ordinary skill in the art a clear understanding of the invention's scope. Throughout the detailed discussion, numerical values and ranges are provided for various implementation aspects. However, these values and ranges are merely illustrative examples and should not limit the applicability of the claims. The article also acknowledges a range of materials suitable for certain aspects of the implementations, emphasizing that these materials are presented as examples and should not restrict the innovation's application.

[027] Referring now to the drawings, these are illustrated in FIG. 1 and 2, the
present invention discloses a hardware-based (embedded in the racket grip)
end-to-end system for racket activity monitoring and correction using sensor data
analysis on the smartphone.

[028] The present unveils an embedded inertial sensor-driven racket 102 spot activities 101 detection system. This novel system strategically integrates inertial sensors into players' rackets, leveraging a sophisticated predictive model on their smartphones to seamlessly identify different shots played by the players. Through real-time data streaming 103 via Wi-Fi from embedded sensors and a microcontroller, the smartphone 105 not only collects the data but also annotates this data for training the predictive model. The trained model not only identifies unknown shots but also provides personalized corrective measures, revolutionizing the way players enhance their racket game-playing skills.
[029] Following an embodiment of the present invention, the method is implemented through a system that carves out a hollow space in the racket handle, preserving the removed material, usually wood. In this hollow space, we place inertial sensors, a microcontroller with Wi-Fi communication modules, a small Li-ion battery 201, a charging circuit 202, and a switch. The combined weight of these components matches that of the removed material, ensuring no added weight or change in the racket's feel for the players.
[030] Following an embodiment of the present invention, the process also involves getting information from the IMU 203 sensors inside the racket on the smartphone when the player makes different shots such as smash, clear, and net shots. To achieve this, the racket's Wi-Fi module connects 204 with the smartphone through an external network access point like a router. This connection facilitates live streaming of data from the racket to the smartphone, including values from all nine axes (3 accelerometers 301 302 303, 3 magnetometers, and 3 gyroscopes 401 402 403). This data is then used to train the predictive model after annotation.
[031] Following an embodiment of the present invention, Wi-Fi communication 104 could include but is not limited to, ZigBee, BLE, or LoRaWAN can be employed depending upon the distance and battery availability. In some cases, the selected communication mechanisms depend upon the communication restriction in certain areas.
[032] Following an embodiment of the present invention, in creating the predictive model, data associated with the IMU sensors are initially annotated as different shots played by the players like Smash, Net Clear etc. This annotated dataset is further divided into training and testing sets to train a classification model using machine learning or deep learning. Precisely, the dataset showcases the profile of the player in the form of IMU data.
[033] Following an embodiment of the present invention, the system also includes an automatic feedback mechanism that helps in suggesting corrective measures to reduce the deviation from the actual position. The corrective instructions and misalignment are for an expert of the chosen racket activity or coach. The activity detection model on the smartphone would also be recalibrated as per the various parameters like height, age, shoulder length and type of racket game of the player.
[034] Following an embodiment of the present invention, the described method and system offer an end-to-end solution for identifying and correcting racket-based activities using data from the sensor module embedded in the It reduces the training requirement from a specialized coach and enhances trust for both individuals and training institutions.
[035] The above-mentioned invention is provided with the preciseness in its real-world applications to provide an end-to-end coach-free method for the detection and correction of racket-based playing activities. The present invention is an effective and reliable method and system. Although the embodiments and claims herein are detailed, they are not exhaustive, and modifications and variations may be made without departing from the scope of the invention as defined in the appended claims. It is also recommended to consult a patent attorney for proper legal advice when preparing patent specifications.
[036] It is to be understood that the description provided above is meant to be illustrative and not limiting. For instance, the discussed embodiments can be utilized in conjunction with one another. Those with expertise in the field will recognize various alternative embodiments after reviewing the description.
[037] The advantages of the current invention have been previously discussed in specific embodiments. It is important to note that these benefits are not considered critical, necessary, or essential features of any particular embodiment. Additionally, they should not be construed as elements or constraints that are crucial for their occurrence or becoming more apparent.
[038] While specific embodiments have been employed to explain the present invention, it is important to note that these embodiments are only illustrative. The invention is not confined to these specific embodiments and can be modified, expanded, and improved in various ways. Such modifications, changes, additions, and enhancements are considered to fall within the scope of the invention. , Claims:1. A system for implementing a racket-based activity recognition and correction method using embedded sensor probes in the racket and learning model running on the smartphone.

2. The system as claimed in claim 1, comprises the construction of a small sensors-probe with IMU sensor, battery, charging module, micro-controller, and communication device, placing the sensor probe in the rod of the racket, collecting data from the IMU while the player is playing any short, the data is transmitted to the smartphone using a communication device, data annotation, training the learning model on the smartphone, real-time testing of the model while player is playing, and automotive feedback collection and auto-correction via vibration or sound.

3. The system as claimed in claim 1, wherein the signal transmission for communication as claimed in claim 2, can include Bluetooth, Wi-Fi, or cellular network based on the distance between the player and smartphone and the used battery in the rod.

4. The system as claimed in claim 1, wherein the selected sensor probe comprises one or more IMU sensors, Li-ion batteries, and communication devices as claimed in claim 2.

5. The system as claimed in claim 4, wherein the constructed sensor probes are located within a predefined location in the rod as per the chosen racket sport.

6. The system as claimed in claim 1, wherein the predictive model is generated by annotating the data associated with racket activities according to a set of shorts played by the player in a particular game, thereby generating a set of player profiles for each of the racket games.

7. The system as claimed in claim 1, involves players' profile variables that consist of the baseline sensory value of a trained person or coach. Deviation from the baseline is estimated during correction based on the initial calibration of height, weight, and the type of racket game played.

8. The system as claimed in claim 1, further comprises a real-time sensor data monitor for recognizing the racket-based sports activity on the smartphone as claimed in claim 2, and for transmitting data associated with the specific sport to a communication network from a sensor probe embedded in the rod of the racket as claimed in claim 3.

9. The system as claimed in claim 2, wherein the sensor data-based activity recognition smartphone is configured to receive the transmitted data, parse it for players profile variable data, and verify its correctness according to a set of trainer or expert profiles.

10. The system as claimed in claim 2, wherein the sensor-based activity recognition is further configured to generate a correctness score based on the processed data deviation from the expert.

11. The system as claimed in claim 1, wherein the system provides an improved solution for recognizing and correcting racket-based sports activity, thereby encouraging coach/trainer-free training and promotion of racket-based activities.