Diagnosis and monitoring of bruxism using earbud motion sensors
Enclosed are embodiments for diagnosis and monitoring of bruxism using earbud motion sensors. In an embodiment, a method comprises: receiving, with at least one processor, a signal derived from a motion sensor in an earbud, wherein the signal is captured while the earbud is inserted in an ear of a user; segmenting, with the at least one processor, the signal into segments; extracting, with the at least one processor, features from the segments; classifying, with the at least one processor, the features; and determining, with the at least one processor, that orofacial activity is predicted based on the classifying.
1 . A method comprising:
receiving, with at least one processor, a signal derived from a motion sensor in an earbud, wherein the signal is captured while the earbud is inserted in an ear of a user, wherein the signal comprises windows of signal samples;
computing an energy of each window of signal samples;
computing a median value of the energies of the windows of signal samples;
determining a portion of the median value as a power threshold; and
adding together windows having an average energy that exceeds the power threshold to reconstruct the signal;
segmenting, with the at least one processor, the signal into segments;
extracting, with the at least one processor, features from the segments;
classifying, the at least one processor, the features; and
determining, with the at least one processor, that orofacial activity is predicted based on the classifying.
2 . The method of claim 1 , wherein the signal is a vibration in an audio frequency band output by an accelerometer in the earbud.
3 . The method of claim 1 , wherein prior to extracting, the signal is pre-processed to remove silent segments.
4 . The method of claim 1 , wherein prior to extracting, the signal is pre-processed to remove segments having energy below a specified threshold.
5 . The method of claim 1 , wherein extracting features from the segments, further comprises:
extracting short-term features using a first sampling window;
extracting mid-term statistical features using a second sampling window longer than the first sampling window; and
long-term averaging the mid-term statistical features using a third sampling window longer than the first and second sampling windows.
6 . The method of claim 5 , wherein the short-term features include time domain and frequency domain features.
7 . The method of claim 1 , wherein the classifying is implemented using a support vector machine (SVM) classifier with a linear kernel.
8 . The method of claim 1 , wherein an amplitude of the signal is normalized as a percent of maximum voluntary clenching (MVC).
9 . The method of claim 1 , wherein the classifying generates classifications in multi-class and binary formats.
10 . A system comprising:
at least one processor;
memory storing instructions, that when executed by the at least one processor, causes the processor to perform operations comprising:
receiving a signal derived from a motion sensor in an earbud, wherein the signal is captured while the earbud is inserted in an ear of a user, and wherein the signal comprises windows of signal samples;
computing an energy of each window of signal samples;
computing a median value of the energies of the windows of signal samples;
determining a portion of the median value as a power threshold; and
adding together windows having an average energy that exceeds the power threshold to reconstruct the signal;
segmenting the signal into segments;
extracting features from the segments;
classifying the features; and
determining that orofacial activity is predicted based on the classifying.
11 . The system of claim 10 , wherein the signal is a vibration in an audio frequency band output by an accelerometer in the earbud.
12 . The system of claim 10 , wherein prior to extracting, the signal is pre-processed to remove silent segments.
13 . The system of claim 10 , wherein prior to extracting, the signal is pre-processed to remove segments having energy below a specified threshold.
14 . The system of claim 10 , wherein extracting features from the segments, and the operations further comprise:
extracting short-term features using a first sampling window;
extracting mid-term statistical features using a second sampling window longer than the first sampling window; and
long-term averaging the mid-term statistical features using a third sampling window longer than the first and second sampling windows.
15 . The system of claim 14 , wherein the short-term features include time domain and frequency domain features.
16 . The system of claim 10 , wherein the classifying is implemented using a support vector machine (SVM) classifier with a linear kernel.
17 . The system of claim 10 , wherein an amplitude of the signal is normalized as a percent of maximum voluntary clenching (MVC).
18 . The system of claim 10 , wherein the classifying generates classifications in multi-class and binary formats.
19 . A method comprising:
receiving, with at least one processor, a signal derived from a motion sensor in an earbud, wherein the signal is captured while the earbud is inserted in an ear of a user;
segmenting, with the at least one processor, the signal into segments;
extracting, with the at least one processor, features from the segments, wherein extracting features from the segments includes:
extracting short-term features using a first sampling window;
extracting mid-term statistical features using a second sampling window longer than the first sampling window; and
long-term averaging the mid-term statistical features using a third sampling window longer than the first and second sampling windows;
classifying, the at least one processor, the short-term features and the long-term averaged mid-term statistical features; and
determining, with the at least one processor, that orofacial activity is predicted based on the classifying.