WO2017092225A1

WO2017092225A1 - Emg-based wearable text input system and method

Info

Publication number: WO2017092225A1
Application number: PCT/CN2016/080732
Authority: WO
Inventors: 伍楷舜; 邹永攀; 叶树锋
Original assignee: 深圳大学
Priority date: 2015-12-04
Filing date: 2016-04-29
Publication date: 2017-06-08
Also published as: CN105511615A; CN105511615B

Abstract

The present invention relates to the field of intelligent identification and discloses an EMG-based wearable text input system. The system comprises a wristband main body and a sensor module. The wristband main body is embedded with an inertia measuring unit (IMU), a processor, a memory, a communication module and an expansion slot. The sensor module is embedded in the expansion slot. The sensor transmits change information of myoelectric bioelectricity to the processor. The IMU records and analyzes motion information of an arm and transmits to the processor. The communication module transmits the analyzed character information. The advantageous effects of the present invention are that a smart wristband is more convenient to wear and more fashionable, the detachable design adds fun to the device, and the reasonable choice of a sensor can increase the precision and also reduce energy consumption.

Description

EMG-based wearable text input system and method

Technical field

The present invention relates to the field of intelligent identification, and in particular, to an EMG-based wearable text input system and method.

Background technique

Today, wearable technology has always been a concern in the international computer academia and industry, but because of the high cost of construction and technical complexity, many related equipment only stays in the conceptual field. With the development of mobile Internet, technological advancement and the introduction of high-performance low-power processing chips, some wearable devices have been conceptualized and commercialized. New wearable devices are constantly coming out, Google, Apple, Microsoft, Sony, and Olin. Bath, Motorola and many other technology companies have begun to explore this new field. "Wearing smart device" is a general term for applying wearable technology to intelligently design and wear wearable devices for daily wear, such as glasses, wristbands, watches, clothing and shoes, which greatly improves the quality of life and work. The efficiency has gradually developed into miniaturization, multi-functionality and strong endurance. According to research, by 2016, the global market for wearable smart devices will reach $6 billion.

Under this trend, many researchers have been digging deep into the potential of wearable technology, exploring the combination of wearable technology and other tools or technologies, such as Google's Google glass, Apple's Apple watch and so on. In contrast, the traditional keyboard is extremely cumbersome and has a single function. There is no effective wearable virtual keyboard application.

Summary of the invention

In order to solve the problems in the prior art, the present invention provides an EMG-based wearable text input system and method, which solves the problem that no wearable device realizes text recognition in the prior art.

The present invention is achieved by the following technical solutions: an EMG-based wearable text input system is designed and manufactured, including a wristband main body and a sensor module, and the wristband main body is embedded with an inertial test unit IMU, a processor, and a memory. a communication module and an expansion slot; the sensor module is embedded in the expansion slot; the sensor transmits electromyographic bioelectrical change information to the processor; and the inertia test unit IMU records motion information of the analysis arm and transmits to The processor; the communication module transmits the analyzed character information.

As a further improvement of the present invention, the sensor module is a myoelectric sensor that senses the myoelectric bioelectrical changes produced by the tapping action, converts them into discrete digital signals and sends them to the processor.

As a further improvement of the present invention, the wristband body has a built-in power module, the power module is a battery or an external charger; and the interface on the wristband body is used for charging or data transmission.

As a further improvement of the present invention, the processor performs feature extraction, self-learning, key motion recognition and matching, and text combination on the signal transmitted by the sensor module and the inertial test unit IMU through the communication mode. The block is sent out.

As a further improvement of the invention: the myoelectric sensor comprises five electrodes, two of which are used to connect the expansion slots on the sides and the other three to be used to sense the myoelectric changes at the bottom.

The invention also provides an EMG-based wearable text input method, comprising the steps of: (S101) receiving data information of a sensor module; (S102) performing feature extraction; (S103) establishing a feature set; (S105) character recognition (S105) Generate text characters.

As a further improvement of the present invention, in the step (S101), according to the training model, the signals sensed by the sensor module and the inertial test unit IMU in each time window are collected, and a data model is established and stored in the memory; The sensor module captures the bioelectrical change signal generated by the electromyogram when the finger hits the paper keyboard; the IMU captures the reading of the acceleration sensor and the gyroscope when the arm moves, and analyzes the movement trajectory of the arm.

As a further improvement of the present invention, in the step (S102), the noise affecting the EMG signal is removed by a filtering algorithm, and then the useful feature signal is extracted; in the step (S103), the latest feature data is performed according to the training model. Processing, establishing a structured feature model; in the step (S105), analyzing the data model, identifying a corresponding button of each action according to the matching algorithm, and determining a character to be input by the user; the step (S106) In the grammar rules, the missing characters are contextually analyzed, then the blur is fixed, and the text information is output.

As a further improvement of the present invention, the EMG-based wearable text input method further includes a self-learning step, which is: according to the latest feature set, combined with the historical feature model, further correcting the data to provide an accurate identification model. The data provides a flexible interface.

As a further improvement of the present invention, the sensor module is a myoelectric sensor that senses a myoelectric bioelectrical change generated by a tapping action, converts it into a discrete digital signal and transmits it to a processor; The test unit IMU records the motion information of the analysis arm and transmits it to the processor; the communication module transmits the analyzed character information.

The beneficial effects of the invention are: the smart wristband is more portable and more fashionable; the detachable design increases the fun of the device, and the reasonable selection of the sensor can improve the accuracy and reduce the energy consumption.

[Description of the Drawings]

1 is a schematic diagram of a demonstration of an EMG-based wearable text input system of the present invention.

2 is a side elevational view of the wristband body of the EMG-based wearable text input system of the present invention.

3 is a developed view of a wristband body of an EMG-based wearable text input system of the present invention.

Figure 4 is a schematic view showing the connection of the myoelectric sensor of the present invention.

FIG. 5 is a schematic structural diagram of an EMG-based wearable text input system according to the present invention.

6 is a schematic flow chart of an EMG-based wearable text input method according to the present invention.

【detailed description】

The invention will now be further described with reference to the drawings and specific embodiments.

An EMG-based wearable text input system includes a wristband body and a sensor module, the wristband body embedding an inertial test unit IMU, a processor, a memory, a communication module, and an expansion slot; the sensor module is embedded in the Expanding the slot; the sensor transmitting the electromyographic bioelectrical change information to the processor; the inertia test unit IMU records the motion information of the analysis arm and transmits to the processor; the communication module analyzes the character Information is transmitted.

The sensor module is a myoelectric sensor that senses the myoelectric bioelectrical changes produced by the tapping action, converts them into discrete digital signals, and sends them to the processor.

The wristband body has a built-in power module, and the power module is a battery or an external charger; the interface on the wristband body is used for charging or data transmission.

The processor performs feature extraction, self-learning, key motion recognition and matching, and text combination on the signals transmitted by the sensor module and the inertial test unit IMU, and then sends the matched characters through the communication module.

The myoelectric sensor contains five electrodes, two of which are used to connect the expansion slots on the sides and the other three to sense the myoelectric changes at the bottom.

In the step (S101), according to the training model, the signals sensed by the sensor module and the inertial test unit IMU in each time window are collected, and a data model is established and stored in the memory; wherein the sensor module captures a finger tap The bioelectrical change signal generated by myoelectricity on the paper keyboard; the IMU captures the readings of the acceleration sensor and the gyroscope when the arm moves, and analyzes the movement trajectory of the arm.

In the step (S102), the noise affecting the EMG signal is removed by the filtering algorithm, and then the useful feature signal is extracted; in the step (S103), the latest feature data is processed according to the training model to establish a structured a feature model; in the step (S105), analyzing the data model, identifying a corresponding button of each action according to the matching algorithm, and determining a character to be input by the user; in the step (S106), according to the grammar rule, Context analysis of missing characters, then blur repair, output text information.

The EMG-based wearable text input method further includes a self-learning step, which is: according to the latest feature set, combined with the historical feature model, further correcting the data, providing accurate data for the recognition model, and providing flexible interface.

The sensor module is a myoelectric sensor that senses a myoelectric bioelectrical change generated by a tapping action, converts it into a discrete digital signal and sends it to a processor; the inertial testing unit IMU records an analysis arm The motion information is transmitted to the processor; the communication module transmits the analyzed character information.

In contrast, the present invention can be combined with other sleep monitoring, stepping and other functions on the same device, providing multiple functions at the same time, and more fashionable, according to successful cases using the myoelectric sensor, such as the MYO control arm ring, The EMG sensor provides accurate EMG electrode change monitoring, even for simple tapping actions. The present invention can use the myoelectric sensor to perceive the tapping action, and combines each tapping action with the character input through comprehensive analysis, as shown in FIG.

In an embodiment, as shown in FIG. 2, FIG. 3, FIG. 4 and FIG. 5, FIG. 2 is a side view of the smart wristband, wherein P101 is a processing module, P102 is a myoelectric sensor; and FIG. 3 is an expanded view of the smart wristband. P201 is the expansion slot of the main wristband (eight in total), P202 is the internal connection point of the expansion slot, P203 is the processing module, P204 is the data transmission bus and power cable; Figure 4 is a detachable myoelectric sensor, P301 It is three necessary sensing electrodes, P302 is two electrodes for connecting to the expansion slot connection point; Figure 5 is the processing module, P401 is the processor, P402 is the memory, P403 is the six-axis inertial test unit (IMU), P404 It is a communication module, and P405 is a USB interface. An EMG-based wearable text input system is a wearable intelligent system, which mainly includes a wristband main body and a myoelectric sensor module. The myoelectric sensor is detachable and can be selectively embedded in the expansion groove of the wristband main body, and embedded. The position and number will affect the recognition accuracy and equipment energy consumption; the myoelectric sensor senses the myoelectric bioelectricity change and transmits it to the processor for processing; the wristband main processing module embeds a six-axis inertial test unit (IMU), which can record the user's arm movement. Data, the arm movement trajectory can be analyzed; the wristband main processing module has embedded processor and memory, and contains various processing algorithms to process the information acquired by the sensor; the wristband main body analyzes the character information through the embedded communication module. Transfer to other electronic devices. An EMG sensor with the same function is optionally embedded in the expansion slot, and the sensor senses the electromyographic bioelectrical changes produced by the tapping action, converts it into discrete digital signals, and sends them to the processor.

The position and number of embedding of the myoelectric sensor will affect the accuracy. The more accurate the position of the sensor covering the myoelectricity, the higher the accuracy of the data. The more the sensor is used, the higher the accuracy, but the more energy is consumed.

The main body of the wristband has a built-in power module. The power supply is provided by the battery. The battery can be charged by a USB interface. The USB interface can be used for charging, and can also be connected to other electronic devices to transmit data. The main body is designed with an expansion slot. The sensor can be embedded and embedded. Data can be transmitted, can also be charged; EMG sensor, built-in battery, can be independently charged; processor, contains a variety of algorithms (such as including "filtering, correction, matching" algorithm), signals sent by myoelectric sensors and IMU sensors Perform "feature extraction - self-learning - key motion recognition and matching - text combination", and finally send the matched characters through the communication module.

The present invention also provides an EMG-based wearable text input method. In one embodiment, the following The step completes the text entry.

S101: During the data collection phase, according to the training model, the signals sensed by the EMG and IMU sensors in each time window are collected, and a data model is established and stored in the memory; wherein the EMG sensor captures the muscles when the finger hits the paper keyboard. The bioelectrical change signal generated by the electric; the IMU sensor captures the readings of the acceleration sensor and the gyroscope when the arm moves, and analyzes the movement trajectory of the arm.

S102: Feature extraction. The data model is acquired from S101, and combined with the data acquired by the IMU, the noise affecting the EMG signal is removed by a filtering algorithm, and then the useful characteristic signal is extracted.

S103: Feature set. According to S102, a useful feature signal is obtained, which is recorded in the feature set, and the feature feature stores the historical feature data and the latest feature data. According to the training model, the latest feature data is processed to establish a structured feature model. Combined with historical feature data, the existing feature models are gradually improved.

S104: Self-learning interface. According to the latest feature set, combined with the historical feature model, the data is further modified to provide accurate data for the recognition model, providing a flexible interface.

S105: Identify the model and the key recognition. The data model provided by the self-learning interface is analyzed, and according to the matching algorithm, the corresponding button of each action is identified, and the character that the user wants to input is determined.

S106: A grammar-based text input interface. According to the grammar rules, the missing characters are contextually analyzed, then blurred and finally combined into words or even sentences.

The above is a further detailed description of the present invention in connection with the specific preferred embodiments, and the specific embodiments of the present invention are not limited to the description. It will be apparent to those skilled in the art that the present invention may be made without departing from the spirit and scope of the invention.

Claims

An EMG-based wearable text input system, comprising: a wristband body and a sensor module, wherein the wristband body is embedded with an inertial test unit IMU, a processor, a memory, a communication module, and an expansion slot; a module embedded in the expansion slot; the sensor transmitting electromyographic bioelectrical change information to the processor; the inertia test unit IMU records motion information of an analysis arm and transmits to the processor; the communication module will The analyzed character information is transmitted.
The EMG-based wearable text input system according to claim 1, wherein the sensor module is a myoelectric sensor, and the myoelectric sensor senses a myoelectric bioelectrical change generated by a tapping action, and converts the same. A discrete digital signal is sent to the processor.
The EMG-based wearable text input system according to claim 1, wherein the wristband body has a power module built in, the power module is a battery or an external charger; and the interface on the wristband body is used for Charging or data transfer.
The EMG-based wearable text input system according to claim 1, wherein the processor performs feature extraction, self-learning, keystroke recognition and matching on signals transmitted by the sensor module and the inertial test unit IMU. After the text is combined, the matched characters are sent out through the communication module.
The EMG-based wearable text input system according to claim 2, wherein the myoelectric sensor comprises five electrodes, two of which are used to connect the expansion slots on the side and the other three are used on the bottom. Perceive changes in myoelectricity.
An EMG-based wearable text input method, comprising: (S101) receiving data information of a sensor module; (S102) performing feature extraction; (S103) establishing a feature set; (S105) character recognition; (S105) Generate a text character.
The EMG-based wearable text input method according to claim 6, wherein in the step (S101), the sensor module and the inertial test unit IMU are perceived in each time window according to the training model. Collected and built a data model stored in the memory; the sensor module captures the bioelectrical change signal generated by the myoelectric force when the finger hits the paper keyboard; the IMU captures the reading of the acceleration sensor and the gyroscope when the arm moves, and the analysis can be obtained. The movement of the arm.
The EMG-based wearable text input method according to claim 6, wherein in the step (S102), the noise affecting the EMG signal is removed by a filtering algorithm, and then the useful feature signal is extracted; (S103), processing the latest feature data according to the training model to establish a structured feature model; in the step (S105), analyzing the data model, and identifying the corresponding button of each action according to the matching algorithm Determining a character to be input by the user; in the step (S106), performing context analysis on the missing character according to the grammar rule, and then blur repairing, Output text information.
The EMG-based wearable text input method according to claim 6, wherein the EMG-based wearable text input method further comprises a self-learning step, specifically: combining history according to the latest feature set The feature model further modifies the data to provide accurate data for the recognition model and provides a flexible interface.
The EMG-based wearable text input method according to claim 6, wherein the sensor module is a myoelectric sensor, and the myoelectric sensor senses a myoelectric bioelectrical change generated by a tapping action, and converts the same. The discrete digital signals are sent to the processor; the inertial test unit IMU records the motion information of the analysis arm and transmits to the processor; the communication module transmits the analyzed character information.