WO2016149831A1 - Systèmes, procédés et dispositifs de reconnaissance d'activité - Google Patents

Systèmes, procédés et dispositifs de reconnaissance d'activité Download PDF

Info

Publication number
WO2016149831A1
WO2016149831A1 PCT/CA2016/050347 CA2016050347W WO2016149831A1 WO 2016149831 A1 WO2016149831 A1 WO 2016149831A1 CA 2016050347 W CA2016050347 W CA 2016050347W WO 2016149831 A1 WO2016149831 A1 WO 2016149831A1
Authority
WO
WIPO (PCT)
Prior art keywords
signals
user
wearable device
signal
activity
Prior art date
Application number
PCT/CA2016/050347
Other languages
English (en)
Inventor
Roshanak Houmanfar
Seyed Ali ETEMAD
Leonard Maceachern
Mark Klibanov
Original Assignee
GestureLogic Inc.
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Priority claimed from US14/669,689 external-priority patent/US10575760B2/en
Application filed by GestureLogic Inc. filed Critical GestureLogic Inc.
Publication of WO2016149831A1 publication Critical patent/WO2016149831A1/fr

Links

Classifications

    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/0002—Remote monitoring of patients using telemetry, e.g. transmission of vital signals via a communication network
    • A61B5/0015—Remote monitoring of patients using telemetry, e.g. transmission of vital signals via a communication network characterised by features of the telemetry system
    • A61B5/0024—Remote monitoring of patients using telemetry, e.g. transmission of vital signals via a communication network characterised by features of the telemetry system for multiple sensor units attached to the patient, e.g. using a body or personal area network
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/05—Detecting, measuring or recording for diagnosis by means of electric currents or magnetic fields; Measuring using microwaves or radio waves 
    • A61B5/053—Measuring electrical impedance or conductance of a portion of the body
    • A61B5/0531—Measuring skin impedance
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/103—Detecting, measuring or recording devices for testing the shape, pattern, colour, size or movement of the body or parts thereof, for diagnostic purposes
    • A61B5/11—Measuring movement of the entire body or parts thereof, e.g. head or hand tremor, mobility of a limb
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/103—Detecting, measuring or recording devices for testing the shape, pattern, colour, size or movement of the body or parts thereof, for diagnostic purposes
    • A61B5/11—Measuring movement of the entire body or parts thereof, e.g. head or hand tremor, mobility of a limb
    • A61B5/1118—Determining activity level
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
    • A61B5/316—Modalities, i.e. specific diagnostic methods
    • A61B5/389—Electromyography [EMG]
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
    • G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
    • G01C21/10—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 by using measurements of speed or acceleration
    • G01C21/12—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 by using measurements of speed or acceleration executed aboard the object being navigated; Dead reckoning
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
    • G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
    • G01C21/10—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 by using measurements of speed or acceleration
    • G01C21/12—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 by using measurements of speed or acceleration executed aboard the object being navigated; Dead reckoning
    • G01C21/16—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 by using measurements of speed or acceleration executed aboard the object being navigated; Dead reckoning by integrating acceleration or speed, i.e. inertial navigation
    • G01C21/166—Mechanical, construction or arrangement details of inertial navigation systems
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H20/00—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance
    • G16H20/30—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance relating to physical therapies or activities, e.g. physiotherapy, acupressure or exercising
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B2560/00—Constructional details of operational features of apparatus; Accessories for medical measuring apparatus
    • A61B2560/02—Operational features
    • A61B2560/0223—Operational features of calibration, e.g. protocols for calibrating sensors
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/0002—Remote monitoring of patients using telemetry, e.g. transmission of vital signals via a communication network
    • A61B5/0015—Remote monitoring of patients using telemetry, e.g. transmission of vital signals via a communication network characterised by features of the telemetry system
    • A61B5/0022—Monitoring a patient using a global network, e.g. telephone networks, internet
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/01—Measuring temperature of body parts ; Diagnostic temperature sensing, e.g. for malignant or inflamed tissue
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/02—Detecting, measuring or recording pulse, heart rate, blood pressure or blood flow; Combined pulse/heart-rate/blood pressure determination; Evaluating a cardiovascular condition not otherwise provided for, e.g. using combinations of techniques provided for in this group with electrocardiography or electroauscultation; Heart catheters for measuring blood pressure
    • A61B5/0205—Simultaneously evaluating both cardiovascular conditions and different types of body conditions, e.g. heart and respiratory condition
    • A61B5/02055—Simultaneously evaluating both cardiovascular condition and temperature
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/02—Detecting, measuring or recording pulse, heart rate, blood pressure or blood flow; Combined pulse/heart-rate/blood pressure determination; Evaluating a cardiovascular condition not otherwise provided for, e.g. using combinations of techniques provided for in this group with electrocardiography or electroauscultation; Heart catheters for measuring blood pressure
    • A61B5/021—Measuring pressure in heart or blood vessels
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/02—Detecting, measuring or recording pulse, heart rate, blood pressure or blood flow; Combined pulse/heart-rate/blood pressure determination; Evaluating a cardiovascular condition not otherwise provided for, e.g. using combinations of techniques provided for in this group with electrocardiography or electroauscultation; Heart catheters for measuring blood pressure
    • A61B5/024—Detecting, measuring or recording pulse rate or heart rate
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/103—Detecting, measuring or recording devices for testing the shape, pattern, colour, size or movement of the body or parts thereof, for diagnostic purposes
    • A61B5/11—Measuring movement of the entire body or parts thereof, e.g. head or hand tremor, mobility of a limb
    • A61B5/1123—Discriminating type of movement, e.g. walking or running
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/22—Ergometry; Measuring muscular strength or the force of a muscular blow
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
    • A61B5/316—Modalities, i.e. specific diagnostic methods
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/48—Other medical applications
    • A61B5/4869—Determining body composition
    • A61B5/4875—Hydration status, fluid retention of the body
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/68—Arrangements of detecting, measuring or recording means, e.g. sensors, in relation to patient
    • A61B5/6801—Arrangements of detecting, measuring or recording means, e.g. sensors, in relation to patient specially adapted to be attached to or worn on the body surface
    • A61B5/6813—Specially adapted to be attached to a specific body part
    • A61B5/6823—Trunk, e.g., chest, back, abdomen, hip
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/68—Arrangements of detecting, measuring or recording means, e.g. sensors, in relation to patient
    • A61B5/6801—Arrangements of detecting, measuring or recording means, e.g. sensors, in relation to patient specially adapted to be attached to or worn on the body surface
    • A61B5/6813—Specially adapted to be attached to a specific body part
    • A61B5/6824—Arm or wrist
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/68—Arrangements of detecting, measuring or recording means, e.g. sensors, in relation to patient
    • A61B5/6801—Arrangements of detecting, measuring or recording means, e.g. sensors, in relation to patient specially adapted to be attached to or worn on the body surface
    • A61B5/6813—Specially adapted to be attached to a specific body part
    • A61B5/6829—Foot or ankle
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/68—Arrangements of detecting, measuring or recording means, e.g. sensors, in relation to patient
    • A61B5/6801—Arrangements of detecting, measuring or recording means, e.g. sensors, in relation to patient specially adapted to be attached to or worn on the body surface
    • A61B5/683—Means for maintaining contact with the body
    • A61B5/6831—Straps, bands or harnesses
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7203—Signal processing specially adapted for physiological signals or for diagnostic purposes for noise prevention, reduction or removal
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7225—Details of analog processing, e.g. isolation amplifier, gain or sensitivity adjustment, filtering, baseline or drift compensation
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
    • A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7271—Specific aspects of physiological measurement analysis
    • A61B5/7275—Determining trends in physiological measurement data; Predicting development of a medical condition based on physiological measurements, e.g. determining a risk factor
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems

Definitions

  • TITLE SYSTEMS, METHODS AND DEVICES FOR ACTIVITY
  • the various embodiments described herein generally relate to systems, methods and devices for activity recognition.
  • the various embodiments described herein relate to recognition of exercise activities using a wearable device.
  • Wearable computing systems are an emerging category of devices. These devices enable users to perform a variety of tasks. For example, users may virtually interact with online accounts; record and/or observe information such as videos, images, and sounds; control other computing systems and other connected appliance; interact with other people; and in some instances monitor the current conditions, state and performance of an individual's body.
  • At least one embodiment described herein provides a method of analyzing activity of a user's body using at least one wearable device positioned on one or more body locations of a user, the method comprising: receiving a plurality of signals from at least one sensor of the at least one wearable device; filtering the plurality of signals to generate a plurality of filtered signals; generating motion data from the plurality of filtered signals; classifying the motion data to identify an activity type.
  • the at least one sensor measures acceleration
  • the plurality of signals comprise acceleration signals.
  • generating the motion data comprises computing a plurality of position signals based on the acceleration signals.
  • computing the plurality of position signals comprises removing low-frequency components and integrating twice to obtain position signals.
  • filtering the plurality of signals comprises low pass filtering each of the plurality of signals.
  • generating the motion data further comprises segmenting quasi-periodic portions in the plurality of position signals.
  • classifying the motion data comprises: modeling, using the processor, a plurality of possible activity classes of the at least one wearable device in a segmented portion; and computing a likeliest class for the segmented portion, out of the plurality of possible activity classes.
  • computing the likeliest class comprises comparing the computed likeliest class with at least one previous class computed by the processor.
  • the modeling is Hidden Markov Modeling and wherein each of the plurality of possible activity classes is modeled using a pre-trained modeler.
  • generating the motion data comprises generating a plurality of feature vectors based on the plurality of filtered signals, wherein each feature vector comprises a plurality of feature values, and wherein each feature vector is associated with a time step of the filtered signals.
  • each feature vector comprises time domain and frequency domain data based on the plurality of filtered signals.
  • each feature vector is computed over a moving time window.
  • generating the motion data further comprises mapping the plurality of feature vectors onto a selected feature space.
  • the mapping is determined by ANOVA computation over a training data set.
  • the predetermined feature space is predetermined by Principal Component Analysis of a training data set.
  • classifying the motion data comprises performing, using the processor, hierarchical classification to compute a likeliest class for a current time step.
  • computing the likeliest class comprises comparing the computed likeliest class with at least one previous class computed by the processor.
  • the at least one sensor further comprises a positioning sensor, wherein the plurality of signals comprises position data from the positioning sensor, and wherein generating the motion data further comprises determining position from the position data.
  • At least one embodiment described herein provides a system for analyzing activity of a user's body using at least one wearable device positioned on one or more body locations of a user, the system comprising: at least one wearable device comprising at least one sensor positionable on the user; a processor operatively coupled to the at least one sensor, the processor configured to: receive a plurality of signals from the at least one sensor of the at least one wearable device; filter the plurality of signals to generate a plurality of filtered signals; generate motion data from the plurality of filtered signals; classify the motion data to identify an activity type.
  • At least one embodiment described herein provides a non-transitory computer readable medium storing computer- executable instructions, which, when executed by a computer processor, cause the computer processor to carry out a method of analyzing activity of a user's body using at least one wearable device positioned on one or more body locations of a user, the method comprising: receiving a plurality of signals from at least one sensor of the at least one wearable device; filtering the plurality of signals to generate a plurality of filtered signals; using a processor, generating motion data from the plurality of filtered signals; classifying the motion data to identify an activity type.
  • FIG. 1 is a block diagram illustrating an example embodiment of a system for activity tracking
  • FIG. 2A is a diagram illustrating an exploded view of an example embodiment of a wearable device that can be used in the system of FIG. 1 ;
  • FIG. 2B is a diagram illustrating various examples of the placement on a user's body of example embodiments of a wearable device
  • FIG. 3 is a block diagram illustrating various components of an example embodiment of a wearable device
  • FIG. 4A is a block diagram illustrating various components of an example embodiment of a signal acquisition unit for a wearable device
  • FIG. 4B is a block diagram illustrating various components of another example embodiment of a signal acquisition unit for a wearable device
  • FIG. 4C is a block diagram illustrating various components of a multiplexer configuration that may be used in an example embodiment of a signal acquisition unit for a wearable device;
  • FIG. 5 is a flowchart illustrating an example embodiment of the processing that may be performed on acquired signals when using the system of FIG. 1 ;
  • FIG. 6 is a flowchart illustrating an example embodiment of a method for activity tracking
  • FIG. 7 is a block diagram illustrating various components used to acquire and process bio-signals in an example embodiment of the activity tracking system of FIG. 1 ;
  • FIGS. 8A and 8B are side and front views, respectively, of a user's leg illustrating an example positioning of a wearable device
  • FIG. 9 illustrates an example process for analyzing activity of a user's body using a wearable device positioned on a user's limb
  • FIG. 10A illustrates another example process for analyzing activity of a user's body using a model-based approach
  • FIG. 10B illustrates an example process for training models for use in the process of FIG. 10A
  • FIG. 10C illustrates an example Hidden Markov Model structure
  • FIG. 1 1A illustrates another example process for analyzing activity of a user's body using a feature-based approach
  • FIG. 1 B illustrates an example process for identifying feature spaces for use in the process of FIG. 1 A
  • FIG. 12 illustrates an example hierarchical classification approach usable with the process of FIG. 1 1A;
  • FIG. 13 illustrates another example classification approach usable with the process of FIG. 1 1A;
  • FIGS. 14A and 14B illustrate example acceleration signals and pre- processed acceleration signals, respectively
  • FIGS. 15A and 15B illustrate example acceleration signals and filtered acceleration signals, respectively
  • FIGS. 16A and 16B illustrate example datasets mapped into projected feature spaces in accordance with the process of FIG. 1 1 A.
  • Coupled can have several different meanings depending in the context in which these terms are used.
  • the terms coupled or coupling can have a mechanical, electrical or communicative connotation.
  • the terms coupled or coupling can indicate that two elements or devices can be directly connected to one another or connected to one another through one or more intermediate elements or devices via an electrical element, electrical signal or a mechanical element depending on the particular context.
  • the term “communicative coupling” may be used to indicate that an element or device can electrically, optically, or wirelessly send data to another element or device as well as receive data from another element or device.
  • any recitation of numerical ranges by endpoints herein includes all numbers and fractions subsumed within that range (e.g. 1 to 5 includes 1 , 1 .5, 2, 2.75, 3, 3.90, 4, and 5). It is also to be understood that all numbers and fractions thereof are presumed to be modified by the term "about” which means a variation of up to a certain amount of the number to which reference is being made if the end result is not significantly changed.
  • Described herein are example embodiments of systems, methods and wearable devices for determining various exercise metrics for an individual user.
  • the various metrics can be determined by capturing movement or position data (e.g., using an accelerometer), or a plurality of electrical signals from a skin surface of the user, or both.
  • Calibration data can be generated based on a subset of the captured electrical signals, and in some cases electromyography signals can be processed using the calibration data to determine at least one biometric.
  • biometric features include muscle intensity, muscle coordination, muscle ratio, muscle fatigue, lactate levels, and hydration.
  • the example embodiments of the systems and methods described herein may be implemented as a combination of hardware or software.
  • the example embodiments described herein may be implemented, at least in part, by using one or more computer programs, executing on one or more programmable devices comprising at least one processing element, and a data storage element (including volatile and nonvolatile memory and/or storage elements).
  • These devices may also have at least one input device (e.g. a pushbutton keyboard, mouse, a touchscreen, and the like), and at least one output device (e.g. a display screen, a printer, a wireless radio, and the like) depending on the nature of the device.
  • a computer readable medium such as, but not limited to, ROM, magnetic disk, optical disc
  • the software program code when read by the programmable device, configures the programmable device to operate in a new, specific and predefined manner in order to perform at least one of the methods described herein.
  • the programs associated with the systems and methods of the embodiments described herein may be capable of being distributed in a computer program product comprising a computer readable medium that bears computer usable instructions for one or more processors.
  • the medium may be provided in various forms, including non- transitory forms such as, but not limited to, one or more diskettes, compact disks, tapes, chips, and magnetic and electronic storage.
  • the medium may be transitory in nature such as, but not limited to, wire-line transmissions, satellite transmissions, internet transmissions (e.g. downloads), media, digital and analog signals, and the like.
  • the computer useable instructions may also be in various formats, including compiled and non-compiled code.
  • Electromyography measures electromyographic signals, which are electrical signals associated with muscle contractions.
  • EMG Electromyography
  • the first method may be referred to as needle electromyography, and is an invasive method that involves inserting electrical probes (needles) into target muscle tissues in order to acquire the EMG signal from the muscle.
  • needle electromyography has been used to measure and analyze muscle contractions and activity; this application has typically been limited to hospital or laboratory settings.
  • the second method is referred to as surface electromyography (sEMG) and involves placing sensors at the skin surface close to target muscles. While less invasive than needle EMG, sEMG has much lower specificity than needle EMG.
  • Surface EMG signals result from the superposition of the sensed motor unit action potentials (MUAPs) from all the muscle fibers within sensing distance of the acquiring electrode.
  • MUAPs sensed motor unit action potentials
  • the MUAPs seen in the sEMG signal may be generated by the same muscle or may originate in different muscles. This can result in signal cross-talk where the sEMG energy from one particular muscle is sensed at an electrode location for a different muscle.
  • Acquired sEMG signals may also contain other undesirable components that are sensed along with the desired muscle activity.
  • the undesirable components may include environmental noise (e.g. 60Hz power line "hum") and/or body-produced signals such as electrical signals from the heart (electrocardiograph (ECG) signals).
  • environmental noise e.g. 60Hz power line "hum”
  • body-produced signals such as electrical signals from the heart (electrocardiograph (ECG) signals).
  • Surface EMG may also be susceptible to "motion artifact” noise which results from the relative motion of the sEMG electrode and the muscle tissue.
  • Motion artifact noise can result from muscle fibres moving underneath the surface of the skin relative to a fixed sensor position during a dynamic contraction (e.g. Bicep curl). This type of motion artifact is independent of the sensor used because the muscle always moves underneath the skin.
  • Motion artifacts may also occur where electrodes are positioned at the surface of the skin and can shift position relative to the muscles being targeted due to the skin's pliability.
  • the "motion artifact" effect may be more pronounced when using electrodes embedded in a wearable device (such as textile electrodes or conductive polymers) as compared with sensors adhered to the surface of the skin. This may affect the signal contribution from target muscles as well as undesirable signal components acquired from nearby non- targeted muscles.
  • the "motion artifact” noise may appear as a low frequency component of the desired sEMG signal, while in other cases it may partially overlap signal components that include the desired frequency content of the sEMG signal.
  • Examples of signal processing methods described in accordance with the teachings herein can be applied to remove the unwanted interference from acquired sEMG signals.
  • Using compressive garments may also mitigate some of the motion artifact noise by reducing the motion of the electrodes in the garment relative to the targeted muscles.
  • Wearable devices and systems incorporating wearable devices have recently been developed for a variety of purposes. Wearable devices can be used for controlling and interacting with multimedia and other electronic devices, as well as for fitness and athletic applications. Described herein are example embodiments of wearable devices, systems incorporating one or more wearable devices, and methods for activity recognition and analysis. The systems, methods and devices described herein can be applied for fitness and athletic applications such as monitoring and analyzing performance, and avoiding injury.
  • the systems, methods and devices described herein can acquire a plurality of electrical signals, for example, from an accelerometer, and analyze those signals to recognize an exercise activity being performed by a user.
  • sEMG signals may also be analyzed in conjunction with the identified exercise activity to provide physiological insights to users.
  • other electrical signals such as skin impedance signals and bio-impedance signals may be acquired.
  • a user's biometric features, as well as other metrics, may be monitored and stored during physical activity and at other times. The biometric features and other metrics may be provided as real-time feedback to a user and/or tracked over time to monitor and illustrate a user's progress.
  • a user's body impedance provides a measure of how well the user's body impedes the flow of electric current.
  • a user's body impedance can be separated into the impedance of the user skin surface, referred to as skin impedance, and the impedance of the user's body below the skin surface, referred to as bio-impedance.
  • EMG signals acquired from a user's skin surface will be affected by the skin impedance.
  • a user's skin impedance generally depends on the condition of the surface of the user's skin. As a result, the user's skin impedance can change as the user performs an activity, for example when the user perspires. ln general, a user's skin impedance will be lower when the skin surface is wet than when the skin surface is dry.
  • Bio-impedance can be affected by a user's body composition, as different tissue types tend to have different levels of resistivity. For example, fat tissue has high resistivity, while tissues with high water content will tend to have low resistivity.
  • a user's bio-impedance can also fluctuate as the user performs an activity, for example due to changes in the user's hydration level, which can affect the user's intra-cellular and extra-cellular water content.
  • the various biometric features and metrics determined for a user may be analyzed to identify trends and alert users to those trends. This may enable users to adjust the activities they are performing, or how they are performing activities in response to the trends. For instance, the signal amplitude and frequency content of acquired EMG signals may be analyzed over time to identify muscle growth or muscle efficiency trends. Similarly, muscle coordination (the sequence in which an individual's muscles are operating) may be analyzed to provide recommendations to improve individual performance. Muscle balance (cross-body) may be tracked and recommendations may be provided to users to adjust their training if muscles appear to be unbalanced. The various biometric features and metrics may also be analyzed to identify an injury risk trend, provide feedback to a user indicating a potential risk of injury, and provide recommendations to avoid the potential risk of injury.
  • Muscle fatigue may be tracked and a fatigue alert may be provided to the user when a fatigue trend is identified.
  • a fatigue alert may be used to prevent overtraining.
  • fatigue may be determined by measuring decreases in mean and median frequencies of EMG signals. For example, this may be done using machine learning algorithms pre-trained on a plurality of collected EMG signals.
  • hydration may be tracked to provide feedback to a user to drink water or other fluids when their body exhibits signs of low hydration.
  • Lactate levels may also be monitored to provide feedback to a user, and may provide a recommendation of reducing activity intensity when lactate levels are identified above a particular threshold.
  • System 100 includes a wearable device 102, one or more remote processing devices 108 and a remote cloud server 110.
  • systems for biometric analysis in accordance with the teachings herein may not require any remote processing devices 108 or remote cloud server 1 10.
  • the processing of the bio- signals acquired from a user may be performed by electronic components included in wearable device 102.
  • Wearable device 102 can be removeably securable to a user skin surface with a skin-facing surface of the wearable device positioned to make contact with the skin surface.
  • Wearable device 102 can generally be manufactured of a fabric, cloth or polymer material suitable for being worn in contact with a user's skin.
  • a portion, or all, of the wearable device 102 can be made of breathable materials to increase comfort while a user is performing an activity.
  • the wearable device 102 may be formed into a garment or form of apparel such as a band, shirts, shorts, and sleeves for example.
  • wearable device 102 may be a compression garment manufactured from a material that is compressive.
  • a compression garment may minimize the impact from "motion artifacts" by reducing the relative movement of the wearable device 102 with respect to a target muscle.
  • wearable device 102 may also include anti-slip components on the skin-facing surface. For example, a silicone grip may be provided on the skin-facing surface of the wearable device 102 to further reduce the potential for motion artifacts.
  • the weave of a compression garment textile can be optimized to minimize motion artifacts by reducing motion in preferred directions of the textile weave.
  • additional compression in addition to the compression elsewhere in the garment can be added to a compression garment in locations having electrodes.
  • wearable device 102 includes a plurality of electrodes 106a-106n positioned to acquire an electrical signal from the user's skin surface.
  • the electrodes 106 can be integrated into the material of the wearable device 102.
  • the electrodes 106 can be affixed or attached to the wearable device 102, e.g., printed, glued, laminated or ironed onto the skin-facing surface.
  • the electrodes 106 may be provided separately from the wearable device 102 as an iron-on component that a user can then apply to the wearable device 102.
  • Various other methods of affixing the electrodes 106 to the wearable device 102 may also be used.
  • Wearable device 102 may also be provided without electrodes 106a to 106n. In such embodiments, wearable device 102 may omit the acquisition and analysis of electromyography and other bio- signals, and may instead provide movement data. In some embodiments, the compression garment may also be omitted, and wearable device 102 may instead be provided as a small module that can be fastened, e.g., to a user's clothing using a clip, a hook and loop fastener, adhesive, or the like.
  • Electrodes 106 may be used with wearable device 102.
  • the electrodes 106 may be metallic, conductive polymers, conductive inks, or conductive textiles for example.
  • Wearable device 102 also includes an electronics module 104 coupled to the plurality of electrodes 106.
  • the electronics module 104 can include a power supply, a controller, a memory, a signal acquisition unit operatively coupled to the controller and to the plurality of electrodes 106, and a wireless communication module operatively coupled to the controller.
  • the signal acquisition unit may include a skin impedance acquisition module and an electromyography acquisition module. In some cases, the signal acquisition unit may also include a bio- impedance acquisition module and/or an analog-to-digital converter. Example embodiments of the wearable device 102 will be described in further detail below, with reference to FIGS. 2 to 4.
  • the electronics module 104 of the wearable device 102 can be communicatively coupled to one or more remote processing devices 108a-108n, e.g., using a wireless communication module (e.g., Bluetooth, IEEE 802.1 1 , etc.).
  • the remote processing devices 108 can be any type of processing device such as a personal computer, a tablet, and a mobile device such as a smartphone or a smartwatch for example.
  • the electronics modules 104 can also be communicatively coupled to the cloud server 1 10 over, for example, a wide area network such as the Internet.
  • Each remote processing device 108 and remote cloud server 1 10 can include a processing unit, a display, a user interface, an interface unit for communicating with other devices, Input/Output (I/O) hardware, a wireless unit (e.g. a radio that communicates using CDMA, GSM, GPRS or Bluetooth protocol according to standards such as IEEE 802.11a, 802.11 b, 802.11g, or 802.1 1 ⁇ ), a power unit and a memory unit.
  • the memory unit can include RAM, ROM, one or more hard drives, one or more flash drives or some other suitable data storage elements such as disk drives, etc.
  • the processing unit controls the operation of the remote processing device 108 or the remote cloud server 110 and can be any suitable processor, controller or digital signal processor that can provide sufficient processing power processor depending on the desired configuration, purposes and requirements of the system 100.
  • the display can be any suitable display that provides visual information.
  • the display can be a cathode ray tube, a flat- screen monitor and the like if the remote processing device 108 or remote cloud server 1 10 is a desktop computer.
  • the display can be a display suitable for a laptop, tablet or handheld device such as an LCD- based display and the like.
  • System 100 can generally be used for tracking biometric features, identifying activity, and measuring other metrics for a user.
  • the activity, biometric features and other metrics may be monitored, stored, and analyzed for the user.
  • aspects of the monitoring, storage and analysis of the different biometric features and other metrics may be performed by the wearable device 102, a remote processing device 108, or the cloud server 110.
  • the cloud server 1 10 may provide additional processing resources not available on the wearable device 102 or the remote processing device 108. For example, some aspects of processing the signals acquired by the electrodes 106 may be delegated to the cloud server 1 0 to conserve power resources on the wearable device 102 or remote processing device 108. in some cases, the cloud server 1 0, wearable device 102 and remote processing device 108 may communicate in real-time to provide timely feedback to a user regarding their current biometrics and any trends identified.
  • one of the controller (on the wearable device 102) and the remote processing device 108 can be configured to capture a plurality of electrical signals including at least one sensor signal using the plurality of electrodes 106.
  • the controller or the remote processing device 108 can generate calibration data based on a subset of the plurality of sensor signals acquired.
  • the acquired electrical signals may include at least one skin impedance signal.
  • the controller or the remote processing device 108 can generate the calibration data based on a subset of the electrical signals that includes the at least one skin impedance signal.
  • the calibration data may be used to adjust the processing of the acquired electromyography signals to account for changes in skin impedance.
  • the acquired sEMG signals may be variable depending on skin condition, accordingly calibrating the processing to respond to changes in skin impedance may account for changes in skin condition that affect the raw EMG signal acquired. For example, as a user begins to perspire, moisture accumulates on the skin surface, changing the skin's effective impedance. This can affect the acquired EMG signal, or it may be used as an indication of exertion. Therefore a "skin impedance signal" can be acquired and used to identify the skin impedance.
  • the skin impedance signal is a form of electrical signal that can be acquired by injecting a probe signal at a first electrode and receiving a resulting signal at a second electrode.
  • skin impedance signals are injected and received through neighboring or proximate electrodes, to maximize the effect of skin impedance.
  • the skin impedance signal can be analyzed periodically or continuously to account for the changes in the acquired EMG signal due to perspiration.
  • bio-impedance signals may also be acquired by wearable device 102.
  • Bio-impedance signals are another form of body impedance, which can be affected by changes in a user's hydration levels, among other factors. Dehydration can cause an increase in the body's electrical resistance, and the bio-impedance signals can be used to analyze an individual's hydration levels.
  • the bio-impedance signal is a form of electrical signal that can be acquired using a tetra-polar electrode configuration. A probe signal can be injected using a first pair of electrodes and a resulting signal can be received at a second pair of electrodes.
  • the bio-impedance signal can be analyzed periodically or continuously to determine biometrics for the user such as changes in hydration levels.
  • bio-impedance signals may be injected and received through cross-body or otherwise distant electrode pairs. This may provide a broader view of the state of the user's body, as compared with a localized view that may be provided if the electrode pairs are close together. Accordingly, using cross-body or otherwise distant electrodes may improve the accuracy of the hydration analysis.
  • EMG signals can be collected using a differential pair of electrodes. That is, two electrodes can be used to acquire a differential signal for a single EMG acquisition channel.
  • EMG signal collection can be paused while skin impedance signals and/or bio- impedance signals are acquired.
  • the signal injection used to measure the skin impedance or bio-impedance may saturate the analog front end components of the EMG acquisition unit. Accordingly, the EMG signal collection can be paused while skin impedance or bio-impedance signals are acquired to avoid interference.
  • system 100 may include a plurality of wearable devices 102 that can be removeably secured to the user's skin surface.
  • each of the wearable devices 102 may communicate with each other or with remote processing device 108 or cloud server 1 0. Using a plurality of wearable devices 102 may better facilitate cross-body or contralateral comparisons of EMG signals and other acquired signals. This may facilitate additional biometric features to be determined from the acquired signals.
  • a user may populate a user profile.
  • the user profile may include user demographic data such as sex, height, weight, age, and other demographic details.
  • a user's demographic data may affect the processing used to determine activities, biometric features or other metrics for that user. For example, determined calorie expenditure may be affected by a user's height and weight, while lactate levels may be affected by a user's sex.
  • the user profile may also track the various activities, biometric features and other metrics determined for a user over time. Accordingly, a user's profile may include a historical record of one or more activities, biometrics or other metrics for that user.
  • the user profile may be stored in various locations.
  • a user's profile can be stored in a memory module of the wearable device 102 or in a remote processing device 108.
  • a user's profile may also be stored in a cloud storage module on cloud server 1 10 in some embodiments.
  • the various components may communicate to update features of the user's profile.
  • a user's profile may be stored in different locations due to storage limitations.
  • the wearable device 102 or remote processing device 108 may store limited portions of a user's profile necessary for determining activities, biometric features and other metrics. Additional portions of the user's profile could then be stored on the remote cloud server 10 for retrieval by (and syncing with) wearable device 102 and remote processing device 108.
  • baseline initialization data may be generated for a user when they first use a wearable device 102.
  • the wearable device 02 or remote processing device 108 may indicate to the user basic activities to perform in an initialization routine.
  • the initialization routine may include the individual performing activities that result in a maximum voluntary contraction (MVC) of target muscles with the wearable device 102 positioned to collect EMG signals from those muscles.
  • MVC maximum voluntary contraction
  • the user may be instructed to perform particular activities with maximum strength, and the acquired EMG signals can be used as a reference for analysis of subsequently acquired signals.
  • muscle strength/power may be initialized through a standard cycling power meter and stored in the user profile for future reference.
  • the initialization process can be repeated for different target muscles and used to generate baseline physiological data.
  • some baseline physiological data may be provided with the wearable device 102 without requiring the user to perform a calibration routine. This baseline physiological data may be pre-generated based on aggregate calibration data from a number of other users. The baseline physiological data also may be adjusted based on demographic data provided in the user's profile. Baseline physiological data can be stored in the user's profile and used to identify changes or trends in the user's physiological state or physiological responses. In some cases, the baseline physiological data may be updated over time to account for changes in the user's physiology, as detected by system 00.
  • the remote cloud server 1 10 may store a unique user profile for each user in a plurality of users. Each unique user profile may be associated with one or more wearable devices 102 used by the corresponding user. In some cases, the unique user profile may be associated with a particular remote processing 108 used by a user. The remote cloud server 1 10 may provide an interactive community for various users of the wearable devices 102 to compare activities, biometric features and other features determined by wearable devices 102.
  • the remote cloud server 110 may also enable longer term trends or patterns to be identified for a user or a collective of users. For example, remote cloud server 110 may analyze activity or biometric data acquired from a particular user over a period of time (e.g., several days, weeks, months or even years) to identify a pattern indicative of potential injury or existing injuries. The remote cloud server 110 may then provide an alert to remote processing device 108 or wearable device 102 associated with that user alerting them to the identified trend. Monitoring and tracking various biometrics and metrics for a user over time can also be used to monitor rehabilitation progress and/or to monitor performance improvements over time. For example, trends can be used to ascertain training goals leading up to a set date, such as a competition date. This trend data can be used to facilitate the athlete "peaking" at the competition date.
  • a set date such as a competition date. This trend data can be used to facilitate the athlete "peaking" at the competition date.
  • FIG. 2A shown therein is an exploded view of an example embodiment of a wearable device 200, which may be an embodiment of wearable device 102.
  • Wearable device 200 is an example of a band-type wearable device that can be worn at various positions on a user's body, as shown in FIG. 2B.
  • wearable device 200 is manufactured of an elastic or compressive material, such as neoprene, elastane or the like, with a plurality of electrodes interwoven into the material 202 of the wearable device.
  • the electrodes may be manufactured of conductive fabrics, to maintain the flexibility of wearable device 200.
  • the material 202 may be an elastane polyester blend and the electrodes may be manufactured of conductive nylon.
  • wearable device 200 may be manufactured of non-compressive material, or without interwoven electrodes.
  • electrodes may be fastened on the interior surface of wearable device 200 using a suitable adhesive, stitching or other fastener.
  • the electrodes can be secured in wearable device 200 by overmolding.
  • the electrodes may be located in specific regions of the wearable device 200 so as to be positioned proximate to target muscles. For example, where wearable device 200 is a thigh band, a set of electrodes may be positioned to overlie the vastus lateralis (VL) muscle of the quadriceps. In some cases, another set of electrodes may be positioned to overlie muscles of the hamstring. In other embodiments where wearable device 200 is intended to be worn over another limb, or another portion of a limb, electrodes may be positioned differently to overlie muscles of interest. In some cases, the electrodes can be configured to acquire an aggregate signal from a group of targeted muscles.
  • VL vastus lateralis
  • wearable device 200 may also include an alignment guide to indicate the optimal orientation of the wearable device 200 when being worn.
  • the electrical signals received using the electrodes in the wearable device 200 may be compared to stored alignment data. Based on this comparison, an alert may be provided to the user to adjust the wearable device 200 for optimal sensor placement.
  • the stored alignment data may be dependent on the particular activity being performed by the user.
  • the acquired electrical signals can be monitored for an expected sequence of muscle activations for a known movement associated with the activity. The acquired electrical signals can be compared with stored alignment data to identify any discrepancies in a received muscle coordination pattern.
  • the amplitude of the acquired electrical signals can be analyzed. If the amplitude is below a predetermined amplitude threshold, it can be determined that the sensors are not proximal to the target muscle(s). In some cases, the frequency content of the received electrical signals can be analyzed to determine if motion artifact signal components are outside expected values. If so, this may indicate that the wearable device 200 is not being worn correctly.
  • the alignment data can be generated by a user performing a baseline initialization routine.
  • the system 100 e.g. wearable device 102 or remote processing device 108 may indicate to the user a sequence of actions that require the user to contract one or more muscles in a particular manner.
  • the electrical signals acquired while the user performs the sequence of actions can be used to generate the alignment data.
  • the alignment data can then be stored in the user profile on the wearable device 102, remote processing device 108 or remote cloud server 1 10.
  • the predetermined amplitude threshold can be determined based on the electrical signals acquired during the training routine and be stored as part of the alignment data.
  • the plurality of electrodes in the wearable device 200 may include a greater number of electrodes than needed. This can be done to provide an even distribution of electrodes throughout the wearable device 200 to minimize the potential for misalignment.
  • the signals received from the plurality of electrodes can be compared with stored alignment data, and a subset of electrodes having desired signal components can be identified for use, based on the comparison.
  • Conductive wires 206 are coupled to the electrodes to convey the acquired signals to circuit board 204.
  • the conductive wires 206 may be conductive textile threads.
  • the conductive wires 206 may be gold-plated to improve conductivity.
  • Circuit board 204 contains the various electronic components of wearable device 200 such as a signal acquisition unit and controller. Electronic components of wearable device 200 will be described in further detail below with regard to F!GS. 3 and 4.
  • the wearable device 200 also includes a battery 208 to power the circuit board 204.
  • one or more user inputs 210 may be provided to allow a user to activate/deactivate the wearable device 200 or adjust parameters of the wearable device.
  • the user input 210 may be a push button or capacitive sensor that can be used to activate the wearable device 200.
  • One or more status indicators such as light emitting diodes (LEDs), or a display, may be provided on outer cover 220 to indicate the status of wearable device 200 to the user.
  • a display integrated into outer cover 220 may also provide additional feedback to a user that may include biometrics or other metrics determined according to the teachings described herein.
  • the wearable device 200 may also include haptic feedback modules to provide haptic feedback to a user regarding the status of the wearable device 200 or alerts regarding the biometrics or other metrics determined for that user.
  • the various electronic components of the wearable device 200 can be contained within a waterproof casing.
  • the waterproof casing can be provided by water-tight casing 216, outer cover 220 and an O-ring 218 between the water-tight casing 216 and the outer cover 220.
  • the waterproof casing can prevent malfunctioning of the electronic components of the wearable device 200 by preventing liquids such as sweat or rain water from coming into contact with the circuit board 204. This may also allow the wearable device 200 to be cleaned after use to improve hygiene.
  • the electronic components in the water-tight casing can be removed from wearable device 200 to ensure that they are not damaged when wearable device 200 is cleaned.
  • wearable device 200 may omit the sleeve, and be contained within water-tight casing 216.
  • water-tight casing 216 may be provided with a fastener for attachment to a user's body or clothing.
  • the fastener may be a clip, an adhesive strip, part of a hook and loop closure, or the like.
  • Wearable device 200 may also include an electronics module seat 214.
  • the seat 214 can be used to join the electronic components of the wearable device to the materials to be worn on the user's skin surface.
  • the seat 214 may be a rubberized ring to minimize the rigidity felt by a user. This may minimize the discomfort felt by a user, and reduce potential interference with normal range of motion.
  • the electronic components of the wearable device may be environmentally sealed and joined to the material 202 of the wearable device 200 using an overmold process. This may also provide a waterproof casing for the electronic components of wearable device 200.
  • FIG. 2B shown therein is a diagram 250 illustrating some examples of locations where a wearable device 260, such as wearable device 200, may be removeably secured to a user's skin surface.
  • wearable devices 260 are shown attached to the user skin surface at the upper arm, lower arm, waist, thigh and lower leg.
  • a skilled person will appreciate that other suitable placements may also be selected for a wearable device 260.
  • the particular placement of a wearable device 260 may be determined based on a particular activity being undertaken by an individual.
  • an individual may select a particular placement based on nearby muscle groups the individual wishes to track. For example, a user may select the thigh or lower leg when targeting different muscles in the legs, and the lower or upper arm when performing activities that target the muscles of the arms.
  • an individual may select a particular placement for a wearable device 260 because it is close to or distant from large muscle groups. For example, a user may select a hip bone as a placement for a wearable device as that region does not have high intensity muscle activity in its immediate vicinity.
  • one or more electrodes in a wearable device 260 may be used as reference electrodes that are not intended to capture EMG signals.
  • a reference electrode may be placed on the user's hip bone, at a location distant from most muscle activity, thereby minimizing the contribution of EMG signals to the electrical signals acquired by the reference electrode.
  • the particular placement of a wearable device 260 may be dependent on the size or shape of that particular wearable device.
  • a wearable device 260 that can be placed around a user's waist may be different from a wearable device 260 that can be placed around a user's calf or forearm.
  • wearable device 300 is an example embodiment of a wearable device 102, 200 or 260 that can be used in activity tracking system 100.
  • wearable device 300 can include additional components not shown in FIG. 3, such as a battery module, an input module or user interface.
  • wearable device 300 includes a plurality of electrodes 306A-306N. The plurality of electrodes 306 are positioned in wearable device 300 to acquire an electrical signal from a user's skin surface when wearable device 300 is secured to the user's skin surface.
  • the electrodes 306 may be exposed on the skin-facing side of the wearable device 300 such that they are in contact with the user skin surface when the wearable device 300 is secured to the user skin surface. In some embodiments, some or all of the electrodes 306 may be capacitively coupled to the user skin surface, in which case they may not make electrical contact with the skin surface (e.g., due to an insulating coating).
  • the electrodes 306 may be passive electrodes. In other embodiments, some or all of the electrodes 306 may be active sensors or electrodes, which contain active electronics that may require power to operate. Examples of active electronics include front-end amplifiers or filters. Given the low signal-to-noise ratio (SNR) often found with sEMG signals, active sensors may be used to increase front-end gain and reject common-mode noise. This can increase SNR of the acquired electrical signals before further processing.
  • SNR signal-to-noise ratio
  • Passive electrodes may be preferred in some applications due to the low-energy associated with their use. This can prolong the battery life of wearable device 300 thereby improving its utility for tracking a user's activity over prolonged periods. Passive electrodes may also be more easily integrated in garments.
  • electrodes may be omitted.
  • wearable device 300 can incorporate a compression garment or compressible garment.
  • a compression garment can be used to reduce the relative movement between the location of the electrodes 306 and muscles being targeted. This can minimize the effect of motion artifacts on the EMG signals acquired.
  • the compression or compressible garment may be omitted.
  • wearable device 300 also includes signal acquisition unit 302.
  • Signal acquisition unit 302 can be used to receive and process, or pre-process, the signals acquired using electrodes 306. In other embodiments, for example when electrodes are not provided, signal acquisition unit 302 may be omitted.
  • Signal acquisition unit 302 can include an EMG acquisition module 310, a skin impedance acquisition module 312, a bio-impedance acquisition module 314 and an analog-to-digital converter 316 (A/D converter).
  • analog-to-digital converter 316 can be implemented using a high resolution, high dynamic range A/D converter.
  • the analog-to-digital converter 316 may sample the electrical signals acquired by electrodes 306 at, for example, 500 to 8000 samples per second and convert the signals to discrete time series data signals.
  • one or more of EMG acquisition module 310, skin impedance acquisition module 312, bio-impedance acquisition module 314 may amplify the analog signal (to increase SNR) before providing the analog signal to A/D 316.
  • EMG acquisition module 310, skin impedance acquisition module 312, bio-impedance acquisition module 314 may provide unfiltered signals to analog-to-digital converter 316.
  • Example embodiments of signal acquisition unit 302 will be described in further detail below with reference to FIG. 4.
  • Signal acquisition unit 302 may be configured to acquire a plurality of analog electrical signals using the plurality of electrodes 306 and to convert the acquired electrical signals in unfiltered form to digital signals using analog-to-digital converter 316.
  • a D converter 316 may sample a continuous time analog EMG signal or other electrical signal and convert them to discrete time samples digital signals. This may provide a high quality digital signal with high levels of both undesirable signal components (e.g., noise) and EMG signals of interest (or skin impedance signal or bio-impedance signals).
  • Various digital signal processing techniques can then be used to extract the desired signal components for analysis in determining biometric features and other metrics.
  • signal acquisition unit 302 may not require A/D converter 316.
  • A/D converter 316 may not be necessary.
  • the EMG acquisition module 310, skin impedance acquisition module 312, and bio-impedance acquisition module 314 may be coupled directly to controller 304.
  • Signal acquisition unit 302 is operatively coupled to controller 304.
  • Controller 304 can be configured to control the signal acquisition unit 302 and to acquire a plurality of electrical signals using the plurality of electrodes 306.
  • controller 304 may also be configured to process the acquired electrical signals to determine at least one biometric feature or other metric.
  • Controller 304 may be, for example, a microcontroller or a microprocessor, such as an ARM-based processor.
  • Controller 304 may be operatively coupled to memory module 318.
  • memory module 318 may include volatile and/or non-volatile memory.
  • Non-volatile memory can be used to store program data, such as executable instructions, and signal data such as raw and/or processed electrical signal data (e.g. EMG data, skin impedance data, bio-impedance data), motion sensor data, calibration data, and operational parameters for example.
  • signal data such as raw and/or processed electrical signal data (e.g. EMG data, skin impedance data, bio-impedance data), motion sensor data, calibration data, and operational parameters for example.
  • volatile memory may be used to store program data and signal data on a temporary basis.
  • Controller 304 may also be operatively coupled to a communication module 308.
  • Communication module 308 may be a wireless communication module used to communication with remote processing device 108 or cloud server 1 10.
  • controller 304 may communicate the unprocessed digital signals received from analog- to-digital converter 316 to remote processing device 108 for further processing.
  • communication module 308 may communicate with remote processing device 108 using various communication protocols such as BluetoothTM Low-Energy (BTLE), ANT+ or IEEE 802.1 1 (WiFiTM).
  • BTLE BluetoothTM Low-Energy
  • ANT+ ANT+
  • WiFiTM IEEE 802.1 1
  • Communication module 308 may also include other interfaces for communicating with remote processing devices 108 and cloud server 1 0.
  • communication module 308 may include ports for wired connections such as a Universal Serial Bus (USB) or other port.
  • wearable device may also include removable storage modules such as an SD card for transferring acquired data to a remote processing device 108.
  • Controller 304 can operate as an interface between the signal acquisition unit and the processing of the acquired signals. In some cases, controller 304 may process the acquired signals itself in real-time or near real-time. In some cases, controller 304 may store the unprocessed signals or transmit processed or unprocessed signals to a remote processing device 108 or cloud server 1 10 for analysis using communication module 308.
  • Controller 304 can be coupled to additional sensor module 320.
  • Motion sensor module 320 may include one or more motion sensors used to acquire data regarding an activity being undertaken.
  • the motion sensor module 320 includes an inertia! measurement unit (IMU) 322 and a Global Navigation Satellite System (GNSS) unit 324, however other sensors may also be used.
  • IMU 322 can include one or more sensors for measuring the position and/or motion of the wearable device.
  • IMU 322 may include sensors such as a gyroscope, accelerometer (e.g., a three- axis accelerometer), magnetometer, orientation sensors (for measuring orientation and/or changes in orientation), angular velocity sensors, or inclination sensors.
  • IMU 322 includes at least an accelerometer.
  • GNSS unit 324 may also include one or more sensors or devices used to determine the GNSS coordinates of the wearable device using one or more GNSS system, such as the Global Positioning System (GPS), Galileo, GLONASS, or the like.
  • GNSS coordinates can be used to track a user's movement or determine location, and supplement the motion data from IMU 322 and the analysis performed by controller 304.
  • GNSS coordinates can also be provided to cloud server 1 10 to facilitate comparisons or communication with other users in close proximity.
  • GNSS coordinates can also be used to determine various metrics related to the user activity, such as weather and other atmospheric information.
  • Wearable device 300 may also include an output module 330 coupled to controller 304.
  • Output module 330 may include one or more output or feedback devices that can provide a user of wearable device 300 with information regarding the status of wearable device 300 as well as feedback on the user's biometrics and other metrics.
  • output module 330 may include one or more status indicators such as an LED, light, or a haptic feedback module.
  • Output module 330 may also include a display such as an OLED or LSC display for providing feedback and indications of status to the user.
  • FIG. 4A shown therein is a block diagram 400 illustrating components in an example embodiment of a signal acquisition sub-unit 402, which may be used in an embodiment of signal acquisition unit 302.
  • Signal acquisition sub-unit 402 illustrates how multiplexers can be used to allow skin impedance signals, bio-impedance signals, and EMG signals to be acquired using the same electrodes.
  • a signal acquisition unit may include two signal acquisition sub-units such as signal acquisition sub-unit 402, each coupled to a separate electrode pair.
  • each signal acquisition sub-unit may have a separate A/D converter, while in other embodiments a single A/D converter can be used to convert the electrical signals acquired by each signal acquisition sub-unit.
  • FIG. 4B Another example embodiment of a signal acquisition unit employing multiplexers will be described with reference to FIG. 4B.
  • Signal acquisition sub-unit 402 is coupled to an electrode pair 406 and a controller 404. As shown in block diagram 400, the signal acquisition unit 402 includes a primary multiplexer 420 and the signal acquisition unit 402 is coupled to electrode pair 406 by primary multiplexer 430. Signal acquisition sub-unit 402 also includes EMG acquisition module 410. Skin impedance acquisition module 412, bio- impedance acquisition module 414 and A/D converter 416. The EMG acquisition module 410 is coupled directly to the primary multiplexer 430. [00114] Signal acquisition sub-unit 402 further includes a secondary multiplexer 432.
  • Each of the skin impedance acquisition module 412 and the bio-impedance acquisition module 414 is coupled to the primary multiplexer 430 by the secondary multiplexer 432.
  • the primary multiplexer 430 and secondary multiplexer 432 can be used to switch the acquisition units that are currently sampling the electrodes of electrode pair 406. This can enable the same electrodes to be used by the EMG acquisition module 410, skin impedance acquisition module 412, and bio-impedance acquisition module 414.
  • FIG. 4B shown therein is block diagram 450 showing components in an example embodiment of a signal acquisition unit 452, which may be an embodiment of signal acquisition unit 302.
  • signal acquisition unit 452 is coupled to 4 electrodes 406A, 406B, 406C and 406D.
  • the electrodes 406 can be positioned to acquire electrical signals from the skin surface.
  • the electrodes may be configured as a first electrode pair 406A and 406B, and a second electrode pair 406C and 406D.
  • Each electrode pair is coupled to a primary switch network 480 (i.e. multiplexer) of the signal acquisition unit 452.
  • the primary switch networks 480A and 480B may be referred to as a primary multiplexer module.
  • the primary multiplexer module can be used to switch between an EMG signal acquisition mode where EMG signals are collected and an impedance signal acquisition modes where impedance signals (i.e. skin impedance or bio-impedance) are collected.
  • signal acquisition unit 452 includes a first EMG acquisition module 41 OA coupled to a first pair of electrodes (406A and 406B) by primary switch network 480A, and a second EMG acquisition module 410B coupled to a second pair of electrodes (406C and 406D) by primary switch network 480B. Accordingly, signal acquisition unit 452 may be used to collect EMG signals using the first electrode pair or the second electrode pair.
  • Signal acquisition unit 452 also includes secondary switch networks 482A and 482B.
  • the secondary switch networks 482A and 482B may be referred to as a secondary multiplexer module.
  • the secondary multiplexer module can be used to switch between a skin impedance acquisition mode and a bio-impedance acquisition mode, using the same set of electrodes 406.
  • the controller (not shown) can transmit multiplexer controller signals to the primary multiplexer module and the secondary multiplexer module to switch the signal type being sampled.
  • the primary multiplexer module can be used to switch between acquiring an EMG signal using EMG acquisition module 410 and one of a skin impedance signal and a bio- impedance signal depending on the state of the secondary multiplexer module.
  • Secondary multiplexer module can be used to switch between acquiring a skin impedance signal using skin impedance acquisition module 412 and a bio-impedance signal using bio-impedance acquisition module 414.
  • signal acquisition unit 452 When signal acquisition unit 452 operates in a skin impedance acquisition mode, a signal can be injected using a first electrode of one of the electrode pairs (e.g. electrode 406A) and then a corresponding skin impedance signal can be received using a second electrode of the electrode pair (e.g. electrode 406B). The injected and received signals can be compared to determine the user skin impedance.
  • signal acquisition unit 452 When signal acquisition unit 452 operates in a bio-impedance acquisition mode, a signal can be injected using the first electrode pair and then a corresponding bio- impedance signal can be received using the second electrode pair. The injected and received signals can then be compared to determine the user bio-impedance
  • the skin impedance acquisition module 412 and bio-impedance acquisition module 414 may be provided in a combined impedance module 460.
  • the impedance module 460 may be provided as multi-channel signal injection and sensing module (as is shown in FIG. 4B).
  • the impedance module 460 may be a single- channel signal injection and sensing module, in which case the secondary multiplexer module may include an additional cascade of multiplexers.
  • A/D converter 416 can switch between the input channels from the EMG acquisition module 410, skin impedance acquisition module 412, and bio-impedance acquisition module 414 and provide the corresponding digital signals to the controller.
  • A/D converter 416 may have sufficient input and output channels to process multiple signals in parallel.
  • multiple A/D converters can be used in parallel to increase the number of channels that can be digitized at the same time or to increase the effective sampling rate.
  • FIG. 4C shown therein is a block diagram 490 illustrating components of an example multiplexer configuration that can be used in an example embodiment of a signal acquisition unit to reduce the number of input channels of the A D converter 416.
  • various components of the signal acquisition unit such as the skin impedance, bio-impedance and EMG acquisition modules have bene excluded for ease of understanding.
  • Block diagram 490 illustrates an alternate configuration of a signal acquisition unit where one or more multiplexers 492 can be used to couple a plurality of N electrodes 406 to an A/D converter having M channels, where M is less than N. This can reduce the cost of the A/D converter used to produce a wearable device such as wearable device 300.
  • the controller 404 and signal acquisition unit can be configured to oversample the electrical signals acquired by the plurality of electrodes 406 to accommodate multiplexing performed by multiplexer 492.
  • a sampling rate (fsampiing) determined using equation (1) can be used to sample the electrical signals acquired by electrodes 406 (using acquisition frequency facquisition). - se l ) f sampling fc acquisition
  • the controller 404 and signal acquisition unit may be configured to sample the electrical signals acquired by the electrodes at eight times the acquisition frequency (i.e., 8 kHz).
  • FIG. 5 shown therein is a process flow 500, illustrating various biometric acquisition and processing steps that may be performed using the described embodiments, such as wearable device 300 or system 100.
  • Flow 500 may be performed using embodiments in which electrodes and a signal acquisition unit are provided.
  • one or more raw analog electrical signal is acquired from one or more electrodes.
  • a plurality of electrical signals will be acquired in this fashion.
  • Each signal can be acquired using one of a plurality of electrodes such as electrodes 106, 306 and 406.
  • each electrical signal may be amplified before being provided to an A/D converter. In other cases, each acquired electrical signal may be provided directly to the A/D converter.
  • each acquired analog signal is converted to a digital signal.
  • an A/D converter such as A/D converters 316 and 416 can be used to sample and convert each analog signal to respective discrete time digital signals.
  • various digital signal processing techniques can be used to filter undesired components from each digital signal.
  • traditional digital signal processing techniques in either the time or frequency domain can be used to filter unwanted noise such as power line hum, motion artifacts, radio frequency etc.
  • time domain processing techniques includes digital filters such as finite impulse response (FIR), infinite impulse response (IIR) and moving average (MA) filters.
  • the acquired EMG signals may be filtered to exclude frequency content outside of a frequency range associated with EMG signals.
  • the acquired EMG signals may be filtered using a band-pass filter tuned to pass a predetermined frequency range associated with EMG signals.
  • the predetermined frequency range may include frequencies between 30Hz and 180Hz.
  • each processed signal can be analyzed to classify the signal and extract signal components of interest and to identify biometric features and other features of an activity being tracked.
  • one or more of the processed electrical signals may be combined with motion sensor signals from motion sensors such as IMU 322 and GNSS 324 for example. Analysis of the various received signals can be used to determine biometric features and other metrics for the user.
  • Additional signal information such as motion or position information (e.g., acquired from the IMU or GNSS unit) and other electrical signals such as skin impedance signals and bio-impedance signals can be analyzed along with the acquired EMG signals to determine biometrics and other metrics for a user.
  • data from the EMG signal and secondary signals can be assessed together.
  • data from the EMG signal and secondary signals can be assessed separately.
  • analysis of the EMG and secondary sources can be conducted concurrently or consequently. In embodiments where the analysis is conducted consequently, results from the analysis of the first parameter can inform the analysis of the second parameter.
  • a processed EMG signal may be analyzed in conjunction with analysis of other signals such as a skin impedance signal and/or a bio-impedance signal.
  • the skin impedance signal can be used to generate calibration data that can be used when analyzing the processed EMG signal.
  • the calibration data may allow the analysis to account for changes in the user's physiology that are not apparent from the EMG signal alone.
  • biometric features may include muscle intensity, muscle coordination, muscle ratio, muscle balance, muscle fatigue, activity intensity, lactate, hydration, muscle efficiency, heart rate, respiratory rate, blood pressure, and breathing volume among others.
  • Other metrics for a user may include location, weather time, distance, calories, Vo2 max, body temperature, glucose levels, acceleration, speed, cadence, activity, pedal position, mechanical power, seat height, inclination, wind velocity, ambient temperature, humidity, air pressure, altitude among others.
  • the particular activity being performed by the user can also be identified based on the acquired signals being analyzed.
  • a comparison between electrical signals received from sensors near identical muscles may be used to detect a possible muscle imbalance in one of the muscles. This can be done by comparing the temporal length and intensities of the received EMG signals for those muscles, in combination with sensor signals related to motion and/or position.
  • a pair of wearable devices 102 may be used to monitor muscle balance while a user is running.
  • Each device may be positioned on the upper thigh of each user's leg with the plurality of electrodes positioned to acquire an EMG signal from the vastus lateralis (VL) muscle of the quadricep complex on both legs.
  • the amplitude of the EMG signals may be assessed in combination with motion signals from the IMUs in each wearable device to determine whether the user's running stride is balanced.
  • Muscle imbalance may be indicative of improper form or an impending injury.
  • Various other features could be extracted from EMG signals to give users valuable feedback on other biometrics.
  • the digital signal processing performed at 530 and the feature extraction and classification performed at 540 can be performed by various components of activity tracking system 100.
  • the digital signal generated at 520 may be provided directly to remote processing device 108 or a cloud server 1 10 without any digital processing being performed by wearable device 102.
  • the signals may be processed at 530 by either remote processing devices 108 or cloud server 1 10.
  • the feature extraction and classification preformed at 540 can also be performed by the remote processing device 108 or the cloud server 1 10.
  • cloud server 1 10 may perform feature analysis in addition to feature analysis performed by a remote processing device 108, for example to identify longer term trends for a user.
  • the controller of the wearable device 102 can perform the signal processing at 530. In some cases, the wearable device 102 can then provide the processed signal data to a remote processing device 108 or cloud server 1 10. In other cases, the controller may also perform some or all of the feature extraction and classification at 540. In embodiments where multiple wearable devices are used to assess biometrics or other metrics, one or more of the multiple wearable devices may perform the signal processing, including feature extraction and classification. [00142] At 550, the biometric features and other metrics identified at 540 can be stored or provided to a user as feedback. In various embodiments the features identified at 540 may be stored in memory module 318 on wearable device 300, in a memory module of remote co processing devices 108, in a cloud storage module of cloud server 110, or in a combination of these storage locations.
  • biometric features may include muscle intensity, muscle coordination, muscle ratio, muscle balance, fatigue, activity intensity, lactate, hydration, muscle efficiency, heart rate, respiratory rate, body temperature, blood pressure, breathing volume, Vo2 max, and glucose levels among others.
  • Other metrics that do not require electromyography may also be acquired, as described herein. Examples of other metrics for a user may include location, weather time, distance, calories, acceleration, speed, cadence, activity, pedal position, mechanical power, seat height, inclination, wind velocity, ambient temperature, humidity, air pressure, altitude among others. In some cases, the particular activity being performed by the user can also be identified based on the acquired signals being analyzed
  • the wearable device or a remote processing device may include a display.
  • the wearable device or remote processing device can be configured to display feedback based on the at least one biometric identified at 540.
  • the feedback provided to a user may be a report or a visualization of the at least one biometric.
  • the remote processing device or wearable device may include a user interface for changing or adjusting the display being provided.
  • the display may include different reports or visualizations, and a user may be able to scroll or browse through the different feedback available.
  • the configuration of the feedback on the display can also be customized based on a user's preferences.
  • the display on the remote processing device or wearable device can be placed into a sleep mode to reduce battery consumption.
  • the display may be activated based on input received from a user.
  • the display can be automatically activated where the processing performed by the wearable device, remote processing device or cloud server generates an alert for the user, such as an indication of fatigue, dehydration, or potential injury.
  • the display may also provide summary information to a user regarding a completed activity.
  • historical records stored on a remote processing device or cloud server can be displayed to a user to illustrate progress trends in the user's activity and physiological response.
  • FIG. 6 shown therein is a flowchart illustrating an example embodiment of a method 600 for biometric tracking.
  • Method 600 can be performed using components of system 100 and wearable devices 102 and 300 in which electrodes and a signal acquisition unit are provided.
  • a plurality of electrical signals including at least one EMG signal, are captured using the electrodes on the wearable device.
  • the signals can be digitized by an A/D converter of the signal acquisition unit and provided to the controller on the wearable device.
  • the signals may be amplified prior to being digitized.
  • the signals may be provided to the controller unfiltered such that all filtering of the acquired signals is performed after digitization.
  • the plurality of electrical signals can include skin impedance signals or bio-impedance signals.
  • skin impedance signals can be acquired using two electrodes positioned on the user's skin surface. For example, a small current signal can be injected into the user's skin surface using a first electrode of an electrode pair (e.g. electrode 406A) and received by a second electrode of the electrode pair (e.g. electrode 406B).
  • the electrodes in the electrode pair can be positioned nearby (e.g. within ⁇ 1 cm) to one another to maximize the conduction that occurs through the user's skin.
  • the amplitude of the signal received by the second electrode may be compared to the amplitude of the injected signal to calculate the impedance of the user skin surface.
  • the phase of the received signal may also be compared across differing frequencies.
  • Bio-impedance signals can be acquired by measuring signals using a tetrapolar electrode configuration.
  • a signal can be injected into the user's body using a first electrode pair (e.g. 406A and 406B).
  • the signal received by a second electrode pair e.g. 406C and 406D
  • a comparison of the injected signal to the received signal can be used to determine bio-impedance for the user.
  • the electrode pairs can be positioned at far apart locations on the user skin surface, such as cross-body or at opposite ends of a limb such that skin conduction in negligible. This may also provide an indication of the user's bio-impedance that is more representative of the hydration levels throughout the user's body, rather than in a localized region.
  • the electrodes may be placed on the same limb, such as a first electrode pair on the front of a user's thigh and a second electrode pair on the back of the user's thigh.
  • an AC injection signal can be generated, e.g.
  • the injection signal can be current limited (e.g. less than 1 mA) to prevent a build-up of capacitance.
  • a multiplexer configuration such as the configuration described above with reference to FIG. 4B, can be used to route the injection signal to a first electrode (in the case of skin impedance sensing) or to a first electrode pair (in the case of bio-impedance sensing).
  • the impedance signal received by the second electrode can be provided as an input to a differential amplifier.
  • the injected signal can be tapped to the differential amplifier.
  • the output of the differential amplifier can be fed into a full wave rectifier module which removes the AC component of the differential amplifier output signal to obtain a root mean square (RMS) value of the output signal.
  • the RMS value can be passed through a low pass filter to obtain a DC voltage level.
  • the DC voltage level can be used to determine the magnitude of the impedance signal.
  • an inphase/quadrature (l/Q) demodulator can be used to determine impedance.
  • An injection signal can be generated and injected to the user skin surface using a first electrode (for skin impedance) or first pair of electrodes (for bio-impedance) as described above.
  • An impedance signal can then be acquired using a second electrode (skin impedance signal) or second pair of electrodes (bio- impedance signal) as described above.
  • the impedance signal can be provided to a differential amplifier, and the differential amplifier output signal can be provided to the inphase/quadrature demodulator.
  • the in- phase and quadrature components of the voltage can then be extracted from the amplified voltage signal.
  • An oscillator can be used to demodulate the amplified voltage signal.
  • the oscillator source signal can be configured to correspond to the generated injection signal. That is, the oscillator in- phase signal can have the same frequency and phase as the injected signal.
  • An oscillator quadrature phase signal can then be generated by delaying the in-phase signal by 90°.
  • the in-phase and quadrature phase oscillator signals can be switched and fed to a mixer.
  • the mixer can combine the in-phase and quadrature phase oscillator signals with the amplified voltage signal to generate in-phase and quadrature phase voltage signals respectively.
  • Each of the in-phase voltage signal and the quadrature voltage signal can be passed through a low pass filter to obtain DC values for the in-phase and quadrature signals. These signals can then be used to determine the real and imaginary components of the impedance.
  • calibration data is generated based on a subset of the electrical signals acquired at 610.
  • the plurality of electrical signals acquired at 610 may include one or more skin impedance signals and/or one or more bio-impedance signals.
  • the calibration data can be generated based on acquired skin impedance signals to account for changes in the EMG signal that may be caused by changes in the conductivity of the user's skin surface.
  • the calibration data may be determined by any one or more of the wearable device, remote processing device or cloud.
  • the calibration data is used to process the EMG signal using the calibration data.
  • the calibration data may account for changes in conductivity of the skin surface that affect the EMG signals acquired at 610.
  • At 640 at least one biometric is determined from the processed EMG signal.
  • the biometric may be one or more of muscle intensity, muscle coordination, muscle ratio and muscle fatigue for targeted muscles.
  • the biometrics may be determined from analyzing the EMG signal in combination with additional sensor information such as motion data or bio-impedance data, for example.
  • biometric Once the at least one biometric has been determined at 640, additional biometrics and other metrics may also be determined in a similar manner as described in methods 500 and 600, for example. The at least one biometric may then be stored or displayed as described above at 550 of method 500.
  • FIG. 7 shown therein is a diagram illustrating various components of system 700 that can be used to acquire and process electrical bio-signals, for example as described with reference to FIG. 6.
  • system 700 includes a wearable device that is generally analogous to wearable devices 200 and 300, and which includes a plurality of electrodes 706a-706n positioned to acquire an electrical signal from a user skin surface when the wearable device is secured to the user skin surface.
  • the electrodes 706 are coupled to a signal acquisition unit 702 and a signal injection unit 762.
  • Signal acquisition unit 702 may correspond generally to the signal acquisition unit described herein with reference to FIGS. 3 and 4.
  • the system 700 may also include a processing module 790.
  • the various components of processing module 790 can be distributed between a wearable device, remote processing device and cloud server in different embodiments.
  • Processing module 790 includes a calibration unit 750, a feature processing module 740, one or more communication module 770, one or more storage modules 780 and a signal injection controller 760.
  • Signal acquisition unit 702 is operatively coupled to the calibration unit 750 and the feature processing module 740.
  • Feature processing module 740 includes a pre-processing unit 742 and a feature identification module 744.
  • Signal acquisition unit 702 may acquire electrical signals using the electrodes 706 and digitize the acquired signals as described herein. These signals may then be provided to the calibration unit 750 and/or the feature processing module 740 for further processing.
  • one or both of the feature processing module 740 and the calibration unit 750 may be provided on the wearable device itself.
  • the feature processing module 740 and the calibration unit 750 may be located on a remote processing device or cloud server, and the digitized signals may be communicated without any processing from the wearable device to the feature processing module 740 and calibration unit 750 using one of the communication modules 770.
  • aspects of the feature processing module 740 and the calibration unit can be distributed among the wearable device, remote processing device and cloud server,
  • signal acquisition unit 702 may acquire digitized skin impedance signals provide these to the calibration unit 750.
  • the calibration unit 750 may generate or update calibration data based on the received signals.
  • the calibration data can then be provided to the feature processing module 740 for use in processing the acquired signals and identifying various biometrics from the acquired signals.
  • the calibration unit 750 may also use pre-processing unit 742 to pre-process the received signals to remove noise and other undesired signal components before generating the calibration data.
  • the pre-processing unit 742 can be used to remove undesirable signal components from the received digital signals. In some embodiments, the pre-processing unit 742 may also be used to calibrate the received digital signals using the calibration data generated by the calibration unit 750.
  • the feature identification module 744 can be used to analyze the received electrical signals (and other sensor signals) to identify at least one biometric for the user. In some cases, the feature identification module 744 can be used to determine a plurality of biometrics and other metrics for the user. The feature identification module 744 may also be used to analyze trends and provide feedback and alerts to a user based on the analysis.
  • Signal injection controller 760 may be used to generate and provide precursor signals used in acquiring a skin impedance signal and/or a bio-impedance signal. Signal injection controller 760 may inject small electrical pulses into particular electrodes 706 using signal injection unit 762. The corresponding electrical signals received at other electrodes 706 can be analyzed to determine features such as skin impedance and bio-impedance, as discussed above.
  • the relative amplitude of the received signals can be compared with the amplitude of the injected signals. This can be used to generate calibration data, which can in turn be used to process the EMG signals and determine various biometrics.
  • hydration may be determined by analyzing a bio-impedance signal acquired from an injected electrical pulse. Hydration may be determined in various ways, such as using existing hydration vs. bio-impedance models, lookup tables, or trained machine learning models.
  • Signal injection controller 760 may also be used to detect when one or more electrodes 706 have become disconnected or are not in contact with a user skin surface. By detecting an open circuit in an electrode 706, system 700 may determine that the electrode has lost contact either with the user skin surface, or has become disconnected from the wearable device. Accordingly, an alert may be provided to the user indicating that the lead has been disconnected.
  • biometric data acquired from electrodes proximate to the user's skin may be supplemented with other data, such as motion or position data.
  • electrodes may not be provided and therefore only motion or position data may be available. Nevertheless, in such embodiments it may still be desirable to recognize or identify a specific activity being performed by the user. Recognition of the specific activity may provide additional insights into the data acquired from the electrodes, or may be used independently to measure other metrics, such as activity type.
  • user activity can be recognized and distinguished from other activities using data from an accelerometer sensor.
  • some of the described embodiments facilitate recognition and distinguishing of activity based on data from a three- dimensional accelerometer sensor of a wearable device.
  • wearable devices Conventionally, distinguishing between different activity types on a wearable device can be challenging, because the sensors are mounted locally and are affected significantly by their orientation as well as adjacent joints and muscles. Furthermore, wearable devices are typically small in size and lightweight, and do not have significant processing capability compared to a laptop or desktop computer, for example.
  • the described embodiments provide activity recognition and are capable of being performed even using the limited processing capability of a wearable device.
  • Two example embodiments are described below, however other variations are possible.
  • a feature-based approach various values and statistical features can be extracted from input signals from the accelerometer, and projected onto a space that is selected to facilitate better segregation of data points that correspond to different activity types. Classifiers can then be used to identify the regions in this projected space in which the data points fall, and thereby distinguish between the different activity types.
  • accelerometer data is collected during different activities, and first pre-processed to reduce the effects of noise.
  • Features are extracted from a moving window of the data, and are projected onto the space that separates the activity classes the most or at least better than unprocessed data. Finally, classifiers are trained and used on this space.
  • a generative model is trained for each activity type. After training, the generative model is capable of capturing the temporal and spatial characteristics of each activity type. Different activity types can be distinguished by identifying similarities between the input data with the generative models.
  • FIGS. 8A and 8B there are illustrated side and front views of a wearable device 800 positioned on the leg of a user.
  • Wearable device 800 is generally analogous to wearable device 200 and 300 as described herein, and includes an IMU with an accelerometer 805. Although shown here as positioned on a leg, in other embodiments, wearable device 800 may be positioned on different limbs or body portions.
  • the accelerometer 805 may be positioned laterally along the limb, so as not to interfere with an ordinary range of motion. In other embodiments, accelerometer 805 may be positioned elsewhere along the limb.
  • a coordinate space for accelerometer 805 is chosen such that a positive x-axis extends in an anterior direction, the positive y-axis extends in an inferior direction and the positive z-axis extends in a lateral direction relative to the user's body.
  • Other coordinate spaces and orientations may also be used with suitable modification of the described embodiments.
  • Process flow 900 may be carried out using a wearable device provided with an accelerometer as described herein.
  • Process flow 900 begins at 910, with a wearable device receiving a plurality of signals from at least one sensor, e.g., accelerometer, of the wearable device.
  • the plurality of signals generally comprises at acceleration signals for at least one axis and generally at least three axes (e.g., X-, y- and z-axis).
  • the plurality of signals may also include gyroscope signals, magnetometer signals, electromyography signals, as described herein.
  • each of the plurality of signals may be initially filtered to generate a plurality of filtered signals.
  • the plurality of signals may be low-pass filtered to remove noise.
  • the plurality of signals may also be high-pass or band-pass filtered to remove undesired signal components. Specific examples of filtering are described with reference to FIGS. 10 and 1 1.
  • a controller of the wearable device generates motion data from the plurality of filtered signals.
  • Motion data includes processed (e.g., filtered) signals received from, for example, an IMU 322, along with derivations from the filtered signals, such as velocity and position signals, mean, standard deviation, maximum, minimum and the like.
  • processed signals e.g., filtered signals received from, for example, an IMU 322, along with derivations from the filtered signals, such as velocity and position signals, mean, standard deviation, maximum, minimum and the like.
  • a different processor such as a remote processing device processor or a cloud server processor, may receive the plurality of filtered signals via wired or wireless communication and may generate the motion data from the plurality of filtered signals.
  • the motion data may be classified to identify activity type. Examples of classification approaches are described with reference to FIGS. 10A and 1 1A.
  • Process flow 1000 is an example embodiment of process flow 900, and may be carried out using a wearable device provided with an accelerometer as described herein.
  • Process flow 1000 is an example of a model-based approach.
  • the plurality of acceleration signals are processed to compute position signals.
  • the acceleration signals may be filtered or smoothed to remove noise before further processing.
  • the position signals may be segmented to identify repeated or quasi-periodic signal patterns which may correspond to repetitions of an activity (e.g., steps, cycling repetitions).
  • HMM Hidden Markov Models
  • Process flow 1000 begins at 1010, with a wearable device receiving a plurality of signals from at least one sensor, e.g., accelerometer, of the wearable device.
  • the plurality of signals includes an acceleration signal for at least one axis and generally signals for at least three axes (e.g., x-, y- and z-axis).
  • the plurality of signals may include other signals, such as gyroscopic signals, magnetometer signals, electromyography signals, and the like.
  • Acceleration signals received from the sensors contain information about the motion being performed.
  • the shape of the acceleration signals over time is characteristic of the motion being performed, and each individual's strategy/technique may be observed in the acceleration signals.
  • the acceleration signals may be affected by environmental noise and local vibrations of the sensor.
  • filtering may be performed at 1020.
  • a moving average filter may be applied to each acceleration signal (for each axis of the accelerometer):
  • acc raw (j) is the signal data collected from the sensors at time j
  • acc(t) is the moving average result for time t
  • N is the window length of the moving average filter.
  • Other types of low-pass filtering may also be performed in addition to, or in lieu of the moving average filter.
  • position signals may be computed at 1030.
  • the DC component of the acceleration signals may be removed:
  • N is the width of the window and should be large enough to include at least one local maximum and one local minimum.
  • N may be set to 1 second, however other window lengths may also be used.
  • An intermediate velocity signal can be obtained for each acceleration signal by integrating the zero mean acceleration data: vel ⁇ t)— I acc zm dt where velit) is the velocity at time t.
  • pos(t) vel zm dt where pos(t) is the position at time t.
  • a discrimination routine may be used to discriminate signal patterns or "segments" that correspond to repetitions within the activity. For example, segments may be discriminated by applying an adaptive thresholding technique. For example, a threshold may be computed as the midway point between a local maximum and a local minimum in the position signal. Alternatively, other linear or non-linear functions of mean, midean, minimum, and maximum can be used. Signal values that exceed the threshold can be identified as belonging to a segment. Signal values that fall below the threshold can be discarded. In other embodiments, the threshold may be pre-specified. In still other embodiments, the discrimination routine may be omitted, although this may diminish performance in some cases.
  • the segmented (or non-segmented, if discrimination is not applied) position signals are input to a plurality of pre-trained generative models that correspond to a plurality of possible activity classes.
  • Each of the plurality of pre-trained generative models is pre-trained using data corresponding to a unique activity class, as described with reference to FIG. 10B.
  • a forward algorithm such as the Viterbi algorithm, can be used to compute the probability that the observation sequence is generated by each of the pre-trained generative models.
  • the model with the largest log likelihood value can be selected.
  • the activity class that corresponds to the selected model can be selected, that is, identified, as the current activity class.
  • error can be reduced by using, for example, a voting procedure based on previously selected activity classes.
  • the selected current selected activity class can be stored in a memory, and compared to previously stored activity classes.
  • the voting procedure may compare 3 previously stored activity classes with the current activity class.
  • FIG. 10B there is illustrated an example process for training models for use in process 1000 of FIG. 10A.
  • Modelling signals allows capturing the main attributes and information of a signal and representing them in a useful and compact manner.
  • Hidden Markov Models a type of generative model, are described by Rabiner, Lawrence R., "A tutorial on Hidden Markov Models and Selected Applications in Speech Recognition," Proceedings of the IEEE 77 (2), IEEE: 257-86 (1989).
  • HMMs are a stochastic modelling approach that characterizes the statistical properties of a signal.
  • a HMM represents a time series signal by an evolving unobservable state variable.
  • the state variable transitions through a discrete set of values, and the probability of these transitions is determined by a state transition matrix.
  • HMMs are trained on a sample dataset of signals. Where segmentation is used, the HMMs are trained on the segmented portions of the signal.
  • FIG. 10C An example of an HMM structure is illustrated in FIG. 10C.
  • the signal is modelled by unobservable states (S1 , S2, ... Sn).
  • S1 , S2, ... Sn unobservable states
  • HMMs are used as the generative models described herein.
  • the HMMs can be trained using a training dataset obtained from multiple users, or over multiple sessions, or both, and which may include motion data such as acceleration and position signals.
  • HMMs are generally capable of capturing the temporal and spatial variations in human motion.
  • each signal segment of interest is modeled by each of the plurality of HMMs, which may be used to distinguish between the activity types of walking, running and cycling.
  • the observations (i.e., inputs) of the HMMs can be the position signals from a three-axis accelerometer. In some embodiments, acceleration signals themselves may also be used as observations. Based on the inputs at each timestep, the HMM either remains in its current state or transitions into another state with a probability.
  • FIG. 10B there is described an example process 1060 for training generative models (e.g., HMM) for use with process 1000 to identify activity classes.
  • generative models e.g., HMM
  • Process 1060 begins at 1062 with the receipt of a sample dataset that can include sampled sensor signals from multiple users, or multiple sessions of users, or both.
  • the signals may be filtered using an approach similar to 1020 of process 1000.
  • position signals can be computed from the filtered signals using an approach similar to 1030 of process 1000.
  • the computed position signals can be analyzed, e.g., by a user, to identify signal segments in the computed position signals.
  • the signal segments can be marked or tagged by a technician user, who may be trained to identify signal segments, and the resulting signal segments stored by the processor.
  • the signal segments may be automatically identified using segmentation algorithms. In the latter case, segments may be identified automatically, and the technician user or end user may be prompted to label or tag a corresponding activity class.
  • the HMM can be updated accordingly.
  • the activity class for each identified signal segment may also be marked or tagged.
  • HMMs may be trained on signal segments corresponding to each known activity class. The process may be repeated at 1072 for each activity class that is desired to be identified. Once there are no further activity classes, training may be completed at 1080 and the HMM parameters stored for each model.
  • Process flow 1 100 is an example embodiment of process flow 900, and may be carried out using a wearable device provided with an accelerometer as described herein.
  • Process flow 1 100 is an example of a feature-based approach.
  • signal values and statistical features can be extracted from the plurality of acceleration signals.
  • Feature selection or feature projection can be utilized to project or map the extracted features into a predetermined feature space, where data for each activity is more readily separable.
  • Predetermined feature spaces may be determined as described with reference to blocks 1 166, 1 168 and 1 170 of a process such as process 1 160 of FIG. 1 1 B. Finally, a classifier is used to determine the region of the space in which the data points are projected, and thereby the likely activity.
  • Process flow 1 100 begins at 1 1 10, with a wearable device receiving a plurality of signals from at least one sensor, e.g. , accelerometer, of the wearable device.
  • the plurality of signals includes an acceleration signal for at least one axis and generally signals for at least three axes (e.g., x-, y- and z-axis).
  • the acceleration signals received from the sensors contain information about the motion being performed.
  • the shape, amplitude and frequency components of the acceleration signals over time is characteristic of the motion being performed, and each individual's strategy/technique may be observed in the acceleration signals.
  • the acceleration signals may be affected by environmental noise and local vibrations of the sensor. To reduce the effects of environmental noise, filtering may be performed at 1 120.
  • FIG. 15A One example of a raw acceleration signal is shown in FIG. 15A.
  • a corresponding filtered acceleration signal is shown in FIG. 15B.
  • Filtering may include, for example, lowpass filtering with cut-off frequency between 7- 10 Hz.
  • a plurality of feature vectors is generated based on the plurality of filtered signals.
  • Each feature vector includes a plurality of feature values representative of some feature of the input signal for some time step of the filtered (or unfiltered) input signal.
  • Each feature vector can include time domain and frequency domain data based on the plurality of filtered signals.
  • each feature vector can include values computed from other time steps, for example within a moving time window, mean, average, maximum or minimum within the window and others.
  • the feature vector may be defined as:
  • f it is the ith feature calculated at time t
  • n is the number of features calculated at each time step.
  • values such as mean, minimum, maximum, range, variance, Fourier transform coefficients of the accelerometer signal (e.g., the first five coefficients) and others may be computed and used as feature values of the feature vector.
  • Signal amplitude may be represented as d.
  • the rate of change in the accelerometer signal value, d, in time window T may be computed and included in the feature vector:
  • FFT Fast Fourier Transformations
  • the generated feature vectors are mapped onto the predetermined feature space determined, e.g., using blocks 1166, 1 168 and 1 170 of process 1 160 of FIG. 1 1 B. Mapping may be performed in several ways, depending on the approach used to predetermine the feature space.
  • the generation of feature vectors at 1 30 and mapping at 1 134 may be merged, by computing only those feature values are required in the predetermined feature space.
  • classification of the feature vectors can be performed at 1 150 using pre-trained classifiers. Classifiers may be trained as described with reference to FIG. 11 B. [00233] Accordingly, as described herein, the projected features of each feature vector can be computed at each time step, and the activity type computed by the hierarchical classifier.
  • the selected activity type for the current time step can be compared to previously stored activity types for previous time steps.
  • error can be reduced by using, for example, a voting procedure based on previously selected activity types.
  • previously stored activity types may be compared to the currently classified activity type. For example, the voting procedure may compare 3 previous activity type predictions.
  • the predicted activity type for each time step is classified separately with the feature-based approach, while each signal portion (e.g., segment consisting of multiple time steps) is classified as a whole in the model-based approach. Regardless of whether segments or individual time samples are assessed, an overall classification of the activity type being performed need not rely solely on the current classified activity type. That is, past classification results can be taken into account and analyzed for a more accurate understanding of the activity. One approach is to average the current prediction and the previous predictions.
  • activity types may be similar, and the feature vectors may be close to classification boundaries, resulting in a prediction sequence that toggles between the two activities. In such cases, the majority of the predictions may not provide a good indication of the original classification. In such cases, a Bayesian approach may be employed in place of a majority voting procedure. The Bayesian approach may aggregate the classification results over multiple time steps or segments. A Naive Bayes classifier, or other classifiers such as Support Vector Machines, K-Nearest Neighbour, Artificial Neural Networks and Hidden Markov Models, can be trained to make the classification. [00237] Referring now to FIG. 1 1 B, there is described an example process 060 for determining feature spaces and training classifiers to identify activity types, for use with process 1 100.
  • Process 1 160 begins at 1 162 with the receipt of a sample dataset that can include sampled sensor signals from multiple users, or multiple sessions of users, or both.
  • the signals may be filtered using an approach similar to 1 120 of process 1 100.
  • position signals can be computed from the filtered signals using an approach similar to 1 130 of process 1 100.
  • a predetermined feature space can be selected " at 1 168 to reduce the dimensions of the feature space, and thereby allow selecting the features that are best suited to distinguishing activity class or activity type.
  • Feature selection and feature projection are two approaches that can be used to reduce the dimension of the feature space and allow selection of features that facilitate classification.
  • the dataset of sample data obtained at 1 162 may be analyzed and feature vectors computed as at 1 130 of process 1 100.
  • the resulting feature vectors can be provided in a matrix D f of feature vectors v f , which includes multiple features for different time samples and different activity classes or types.
  • An output matrix L f may be provided, which contains an activity class or activity type that is known a priori for each feature vector v f in the sample dataset D f .
  • ANOVA Analysis of Variance
  • ANOVA can determine the most informative (e.g., distinguishing) features of a feature set.
  • ANOVA determines the probability that a particular feature in a plurality of feature vectors (e.g., mean) is generated from the same distribution. The smaller this probability value is, the more likely it is the feature under consideration originates from a different activity type than the reference feature(s) for a particular activity type.
  • an individual feature e.g., mean
  • F t is the matrix of all the available samples of the feature ft in the training set
  • N is the number of available samples
  • p ⁇ is the probability value calculated for each feature separately using ANOVA
  • F s are the features with the smallest probability p.
  • ANOVA can be applied to the different activity types (or classes) simultaneously.
  • the features identified using ANOVA analysis can be determined as the best distinguishing features for the various activity types.
  • other features subsequently can be discarded or not computed at all. For example, if ANOVA analysis identifies mean, variance, first and third Fourier transform coefficients as distinguishing features, then these may form the common feature space used for classification. Conversely, the second Fourier transform coefficient may be discarded and other features, such as maximum or minimum values, need not be computed at all.
  • FIG. 16A illustrates an example dataset mapped into a projected feature space created using an ANOVA approach. Specifically, feature vectors from the example dataset were projected into a three-dimensional space derived using ANOVA. The activity type of each feature vector is identified by the marker in the plot 1600. Three different activity types were used, with three different markers applied in plot 1600 (i.e., square, triangle and hexagon). The separability of the data in the projected feature space can be seen clearly.
  • PCA Principal Component Analysis
  • eigenvalues and eigenvectors are computed of a covariance matrix for a dataset.
  • the eigenvectors are sorted based on eigenvalues in a descending manner.
  • Each eigenvector can form the basis of a new dimension, onto which the dataset can be projected.
  • Those dimensions which exhibit the most variant directions (e.g., the largest eigenvalues) in the dataset can be selected as the bases of a projected space. Other dimensions can be discarded to reduce data needs and computational complexity.
  • D f D f sorted(y) where D f is the mean of data set D f , D f is the projection of O f onto the new space and sortedQ ) contains the sorted eigenvectors such that the eigenvectors corresponding to higher eigenvalues are ordered first.
  • FIG. 16B illustrates an example dataset mapped into a projected feature space created using a PCA approach.
  • ANOVA feature vectors from the example dataset were projected into a three-dimensional space derived using PCA.
  • the activity type of each feature vector is identified by the marker in the plot 1600.
  • Three different activity types were used, with three different markers applied in plot 1600 (i.e., square, triangle and hexagon). The separability of the data in the projected feature space can be seen clearly.
  • the feature vectors may be mapped onto the determined feature space at 1170.
  • classifiers may be trained for distinguishing activity types in the feature space at 1 172.
  • a hierarchical classification approach can be used as described above.
  • each classifier can be trained to compare a single activity type to all remaining activity types.
  • the preferred feature space for this binary comparison between a first activity type and remaining activity types is found for the training dataset.
  • the process may be repeated at 1 174, until enough classifiers have been trained to classify each activity type, whereupon the process may complete at 1 180.
  • a binary classifier is trained to separate one action from the remaining actions in the trained feature space.
  • the sample data corresponding to the first activity type is removed from the data set, and additional feature spaces are identified for additional activity types in similar fashion.
  • the procedure is repeated until the optimal feature space to separate the second activity from the remaining activities is found in the same manner, and the second layer's classifier is trained on this feature space. This procedure continues until a classifier has been trained for all for all activity types.
  • FIG. 12 illustrates an example hierarchical classification approach. As shown, a first comparison is made at node 1202, which classifies between a first activity type 1205 and remaining activity types 1210. Next, a second comparison is made between remaining activity types 1210 and activity type 1215. Additional comparisons can be made sequentially thereafter.
  • SVM Support Vector Machines
  • NB Naive Bayes
  • SVMs construct a half space in any dimension n, which can be used for the purpose of classification, or regression.
  • a soft margin is employed so that in the absence of a hyperplane that can perfectly separate the data, the classifier will choose a hyperplane that maximizes the possibility of separating the data while still minimizing error.
  • the matrix D f includes the projected features and L ⁇ - includes the corresponding activity types of the data suitable for the binary classification at each level of the hierarchical classifier.
  • the SVM is trained at each level of the hierarchical classifier using a dataset, to separate one activity type from all of the remaining activity types.
  • a Naive Bayes classifier at each step, one activity type is compared to all remaining activity types.
  • the Na ' ive Bayes classifier is trained on the identified feature space using the training data (e.g., at each level) to separate a particular activity type from the remaining activity types in the corresponding feature space.
  • a Naive Bayes classifier operates on the assumption that each input feature, /, is independent of the others:
  • FIG. 13 illustrates a schematic model of a Naive Bayes classifier.
  • K-NN K-Nearest Neighbour
  • the described example embodiments have been validated using experimental data.
  • a ten-fold cross validation was performed in which 90% of a sample data set was used to train feature-based and model-based processors (i.e., "Cross Validation”). The remaining 10% of the data set was used to test the trained processors.
  • sample data for multiple test subjects was used to train, and one test subject was left out of the training dataset and used to verify the trained processors (i.e., "Leave One Out").

Landscapes

  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Remote Sensing (AREA)
  • Physics & Mathematics (AREA)
  • Medical Informatics (AREA)
  • Public Health (AREA)
  • General Health & Medical Sciences (AREA)
  • Biophysics (AREA)
  • Molecular Biology (AREA)
  • Veterinary Medicine (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Pathology (AREA)
  • Biomedical Technology (AREA)
  • Surgery (AREA)
  • Animal Behavior & Ethology (AREA)
  • Oral & Maxillofacial Surgery (AREA)
  • Automation & Control Theory (AREA)
  • General Physics & Mathematics (AREA)
  • Physiology (AREA)
  • Dentistry (AREA)
  • Physical Education & Sports Medicine (AREA)
  • Epidemiology (AREA)
  • Primary Health Care (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Dermatology (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Radiology & Medical Imaging (AREA)
  • Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)

Abstract

L'invention concerne des systèmes, des procédés et des dispositifs permettant de reconnaître l'activité d'un utilisateur à l'aide de données provenant d'un accéléromètre ou d'autres capteurs. L'invention concerne une approche « basée sur des caractéristiques » et des approches « basées sur un modèle », dans une approche basée sur les caractéristiques, diverses valeurs sont extraites depuis des signaux d'entrée et projetées sur un espace qui est choisi pour faciliter une meilleure répartition des points de données. Des dispositifs de classification permettent d'identifier les régions dans l'espace projeté dans lesquelles les points de données sont situés afin de procéder à une distinction entre les différents types d'activités. Dans une approche « basée sur un modèle », un modèle génératif est formé pour chaque type d'activité. Différents types d'activités peuvent être distingués par identification de similitudes entre les données d'entrée grâce aux modèles génératifs.
PCT/CA2016/050347 2015-03-26 2016-03-24 Systèmes, procédés et dispositifs de reconnaissance d'activité WO2016149831A1 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
US14/669,689 2015-03-26
US14/669,689 US10575760B2 (en) 2014-03-26 2015-03-26 Systems, methods and devices for activity recognition

Publications (1)

Publication Number Publication Date
WO2016149831A1 true WO2016149831A1 (fr) 2016-09-29

Family

ID=56979285

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CA2016/050347 WO2016149831A1 (fr) 2015-03-26 2016-03-24 Systèmes, procédés et dispositifs de reconnaissance d'activité

Country Status (1)

Country Link
WO (1) WO2016149831A1 (fr)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109453501A (zh) * 2018-10-25 2019-03-12 惠州市德赛工业研究院有限公司 球类训练数据处理方法、设备
EP3695781A4 (fr) * 2017-10-13 2021-03-10 Sony Corporation Dispositif de mesure d'informations biologiques et système de mesure d'informations biologiques
US11657281B2 (en) 2019-03-15 2023-05-23 Hinge Health, Inc. Method and system for symmetric recognition of handed activities

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CA2818020A1 (fr) * 2011-06-10 2012-12-13 Aliphcom Modeles de profil de mouvement et langues de mouvement pour dispositifs a porter
US20150272482A1 (en) * 2014-03-26 2015-10-01 GestureLogic Inc. Systems, methods and devices for activity recognition

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CA2818020A1 (fr) * 2011-06-10 2012-12-13 Aliphcom Modeles de profil de mouvement et langues de mouvement pour dispositifs a porter
US20150272482A1 (en) * 2014-03-26 2015-10-01 GestureLogic Inc. Systems, methods and devices for activity recognition
US20150272483A1 (en) * 2014-03-26 2015-10-01 GestureLogic Inc. Systems, methods and devices for exercise and activity metric computation

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
LIN, CHIA-HUA.: "A Real-Time Human Posture Classifier and Fall-Detector.", ELECTRONIC THESIS OR DISSERTATION., 2014, pages 7 - 8, 6, 12-13, 19-21, 26-26, 32, 49, 66, XP055316828 *
QUWAIDER, MUHANNAD ET AL.: "Body posture identification using hidden Markov model with a wearable sensor network.", BODYNETS, 2008, XP055316829 *

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP3695781A4 (fr) * 2017-10-13 2021-03-10 Sony Corporation Dispositif de mesure d'informations biologiques et système de mesure d'informations biologiques
US11445932B2 (en) 2017-10-13 2022-09-20 Sony Corporation Biological information measurement device and biological information measurement system
CN109453501A (zh) * 2018-10-25 2019-03-12 惠州市德赛工业研究院有限公司 球类训练数据处理方法、设备
US11657281B2 (en) 2019-03-15 2023-05-23 Hinge Health, Inc. Method and system for symmetric recognition of handed activities

Similar Documents

Publication Publication Date Title
US10575760B2 (en) Systems, methods and devices for activity recognition
US9782104B2 (en) Systems, methods and devices for acquiring and processing physiological signals
WO2016149832A1 (fr) Systèmes, procédés et dispositifs de calcul de mesure d'exercice et d'activité
US10986465B2 (en) Automated detection and configuration of wearable devices based on on-body status, location, and/or orientation
Sztyler et al. Position-aware activity recognition with wearable devices
Beach et al. An ultra low power personalizable wrist worn ECG monitor integrated with IoT infrastructure
US9770179B2 (en) System, method and device for detecting heart rate
EP2678757B1 (fr) Système de reconnaissance gestuelle
US20150057506A1 (en) Arrayed electrodes in a wearable device for determining physiological characteristics
WO2016149829A1 (fr) Systèmes, procédés et dispositifs d'acquisition et de traitement de signaux physiologiques
WO2017079828A1 (fr) Systèmes et procédés pour acquisition et analyse de données de santé
CN104138260B (zh) 一种利用svm分类器的睡眠姿势多分类识别方法
CN110035691A (zh) 用于测量睡眠呼吸暂停的方法和设备
US10368765B2 (en) Wearable apparatus for ECG signal acquisition
CN112057059A (zh) 基于多模态生理数据的心理压力智能采集测试分析系统
CN107847146B (zh) 在ecg数据上自动标记活动的方法和系统
WO2016149831A1 (fr) Systèmes, procédés et dispositifs de reconnaissance d'activité
Altini et al. Cardiorespiratory fitness estimation using wearable sensors: Laboratory and free-living analysis of context-specific submaximal heart rates
Leutheuser et al. Textile integrated wearable technologies for sports and medical applications
Shaji et al. Real-time processing and analysis for activity classification to enhance wearable wireless ECG
WO2023037359A1 (fr) Procédé et système d'analyse de signaux pendant un exercice
WO2016149830A1 (fr) Système, procédé et dispositif pour détecter une fréquence cardiaque
US20230018876A1 (en) Wearable Nystagmus Detection Devices and Methods for Using the Same
GB2605351A (en) System and method for monitoring muscle performance and providing real-time dynamic advice
Jin et al. VTMonitor: Tidal Volume Estimation Using Earbuds

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 16767598

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 16767598

Country of ref document: EP

Kind code of ref document: A1