Determining energy expenditure using a wearable device
A method and system for determining energy expenditure (EE) are disclosed. The method comprises determining a plurality of sensor streams using a plurality of detected physiological signals and processing the plurality of sensor streams to determine the EE. The system includes at least one sensor to detect a plurality of physiological signals, a processor coupled to the at least one sensor, and a memory device coupled to the processor, wherein the memory device includes an application that, when executed by the processor, causes the processor to determine a plurality of sensor streams using a plurality of detected physiological signals and to process the plurality of sensor streams to determine the EE.
1. A method for determining energy expenditure (EE) of a user/patient, the method comprising:
detecting a plurality of physiological signals using a wearable sensor device, wherein the plurality of detected physiological signals are detected by a wearable device and include any of an electrocardiogram (ECG) signal and an acceleration signal;
converting the plurality of physiological signals into a plurality of sensor streams, wherein the plurality of sensor streams include any of body impedance, heart rate (HR), basal heart rate (HR b ), heart rate variability (HRV), and signal magnitude area (SMA);
a preprocessing step for eliminating wearable device off instances using body impedance values;
extracting features from each of the plurality of sensor streams, and determining a feature vector using extracted features and demographic information of the user/patient; and
determining the EE by performing machine learning using the determined feature vector and a regression model, wherein the regression model corresponding to a specific activity state is selected based on value of normalized heart rate (HR nu ).
2. The method of claim 1 , wherein the processing step is performed by any of a wearable device, an external device, a relay/cloud processor, a smartphone device, and a cloud computing system.
3. The method of claim 1 , wherein the preprocessing step further comprises any of:
low-pass filtering the plurality of sensor streams; and normalizing the plurality of sensor streams, and
normalizing the plurality of sensor streams, wherein the normalized heart rate (HR nu ) is determined as any of a function of basal heart rate (HR b ), a function of maximal allowable heart rate (HR max ), and a fraction of heart rate reserve (HRR).
4. The method of claim 1 , wherein the feature vector includes user/patient information related features that include any of height, weight, gender, age; heart rate related features that include any of heart rate (HR), normalized heart rate, and heart rate variability (HRV); and activity features that include signal magnitude area (SMA).
5. The method of claim 1 , wherein the regression model is a piecewise linear regression model with a break point, wherein the break point is a function of a normalized heart rate (HR nu ) and is determined as a fraction of heart rate reserve (HRR), wherein the break point splits the determination of the EE into two separate linear regressions.
6. The method of claim 1 , wherein the regression model is a piecewise linear regression model with two break points, wherein the two break points are a function of a normalized heart rate (HR nu ) and are each determined as a fraction of heart rate reserve (HRR), wherein the two break points split the determination of the EE into three separate linear regressions.
7. The method of claim 1 , wherein the regression model is a nonlinear regression that does not include a break point.
8. The method of claim 1 , wherein to determine the EE comprises the determination of any of an energy expenditure (EE) rate, total daily energy expenditure (TEE), basal energy expenditure (BEE), and a physical activity level (PAL) as the ratio of TEE/BEE.
9. A system for determining energy expenditure (EE) of a user/patient, the system comprising at least one sensor to detect a plurality of physiological signals, a processor coupled to the at least one sensor, and a memory device coupled to the processor, wherein the memory device includes an application that, when executed by the processor, causes the processor to:
convert the plurality of physiological signals detected using the sensor into a plurality of sensor streams,
wherein the plurality of physiological signals include any of an electrocardiogram (ECG) signal and an acceleration signal and
wherein the plurality of sensor streams include any of body impedance, heart rate (HR), basal heart rate (HR b ), heart rate variability (HRV), and signal magnitude area (SMA);
extract features from each of the plurality of sensor streams, and determine a feature vector using extracted features and demographic information of the user/patient; and
determine the EE by performing machine learning using the determined feature vector and a regression model, wherein the regression model corresponding to a specific activity state is selected based on value of normalized heart rate (HR nu ).
wherein the system further comprises a preprocessing module, wherein the processing module eliminates the off instances using body impedance values prior to feature extraction and EE determination.
10. The system of claim 9 , wherein the preprocessing module further comprises any of:
low-pass filtering the plurality of sensor streams; and normalize the plurality of sensor streams; and
normalizing the plurality of sensor streams, wherein the normalized heart rate (HR nu ) is determined as any of a function of basal heart rate (HR b ), a function of maximal allowable heart rate (HR max ), and a fraction of heart rate reserve (HRR).
11. The system of claim 9 , wherein the feature vector includes user/patient information related features that include any of height, weight, gender, age; heart rate related features that include any of heart rate (HR), normalized heart rate, and heart rate variability (HRV); and activity features that include signal magnitude area (SMA).
12. The system of claim 9 , wherein the regression model is a piecewise linear regression model with a break point, wherein the break point is a function of a normalized heart rate (HR nu ) and is determined as a fraction of heart rate reserve (HRR), wherein the break point splits the determination of the EE into two separate linear regressions.
13. The system of claim 9 , wherein the regression model is a piecewise linear regression model with two break points, wherein the two break points are a function of a normalized heart rate (HR nu ) and are each determined as a fraction of heart rate reserve (HRR), wherein the two break points split the determination of the EE into three separate linear regressions.
14. The system of claim 9 , wherein the regression model is a nonlinear regression that does not include a break point.