Powered prosthetic devices using EMG-based locomotion state classifier

• US 9,649,207 B2
• Filed: 07/07/2014
• Issued: 05/16/2017
• Est. Priority Date: 07/12/2013
• Status: Active Grant

• Current Assignee: University of Oregon
• Original Assignee: University of Oregon
• Inventors: Hahn, Michael, Joshi, Deepak
• Primary Examiner(s): Willse, David H
• Application Number:
  US14/903,009
• Publication Number:
  US 20160158033A1
• Time in Patent Office: 1,044 Days
• Field of Search: None
• CPC Class Codes: A61B 5/1123 Discriminating type of movement, e.g. walking or running A61B5/1116, A61B5/112 take precedence
• CPC Class Codes: A61B 5/1126 using a particular sensing technique
• CPC Class Codes: A61B 5/389 Electromyography [EMG]
• CPC Class Codes: A61B 5/7264 Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems neural networks per se G06N3/00; expert systems per se G06N5/00
• CPC Class Codes: A61F 2/60 Artificial legs or feet or parts thereof
• CPC Class Codes: A61F 2/72 Bioelectric control, e.g. myoelectric
• CPC Class Codes: A61F 2002/704 computer-controlled, e.g. robotic control
• CPC Class Codes: G16H 50/20 for computer-aided diagnosis, e.g. based on medical expert systems

Abstract

Real-time control of a prosthetic device using EMG-based locomotion state classification computes a histogram [208] of a time-frequency spectrogram [206] of the EMG data [200] sampled from muscles, classifies the histogram using if-else rules [212] as representing a locomotion steady state [216] or locomotion transition state [214] in the prosthetic device, and controls the prosthetic device using the computed transitions between locomotion modes. The classification may be based on a comparison of feature values [210] derived from the histogram and stored feature values derived from histograms of known locomotion states. Alternatively, the classification may be based on matching scores calculated from the histogram and stored histograms of known locomotion states. The classifying preferably is performed using hierarchical if-else fuzzy classification rules [212], and may further include using a prior locomotion state and a state diagram specifying constraints on locomotion states accessible from other locomotion states.

Claims

1. A method for real-time control of a prosthetic device using EMG-based locomotion state classification, the method comprising:

sampling surface EMG signals from muscles to produce EMG data;

computing a histogram of a time-frequency spectrogram of the EMG data;

computing a feature value of the histogram of the time-frequency spectrogram of the EMG data;

comparing the feature value with stored feature values for locomotion steady states and locomotion transition states;

classifying using if-else rules the feature value as representing a locomotion steady state or locomotion transition state in the prosthetic device;

controlling the prosthetic device using the computed transitions between locomotion modes.

2. The method of claim 1 wherein the time-frequency spectrogram of the EMG data is a 2D spectrogram.

3. The method of claim 1 wherein the feature value is skewness or kurtosis.

4. The method of claim 1 wherein the time-frequency spectrogram of the EMG data comprises energy values assigned to a grid of time bins and frequency bins for EMG data in a time window.

5. The method of claim 4 wherein the energy values are quantized values corresponding to fractions of a maximum energy.

6. The method of claim 5 wherein the energy values are binary values where a value of 1 represents an energy exceeding a predetermined threshold energy and a value of 0 represents an energy not exceeding the predetermined threshold energy.

7. The method of claim 1 wherein the histogram comprises a number of occurrences of maximum energy for frequency bins in the time-frequency spectrogram.

8. The method of claim 1 wherein classifying comprises hierarchical if-else fuzzy classification rules.

9. The method of claim 8 wherein classifying comprises a main classification between locomotion transition state and locomotion steady state, and a subclassification between locomotion transition states and locomotion steady states.

10. The method of claim 1 wherein classifying comprises using a prior locomotion state and a state diagram specifying constraints on locomotion states accessible from other locomotion states.

11. The method of claim 1 wherein the locomotion steady states comprise level walking, stair ascent/descent, and ramp ascent/descent, and wherein the locomotion transition states comprise transitions between the locomotion steady states.

12. A method for real-time control of a prosthetic device using EMG-based locomotion state classification, the method comprising:

sampling surface EMG signals from muscles to produce EMG data;

computing a histogram of a time-frequency spectrogram of the EMG data;

computing matching scores between the histogram of the time-frequency spectrogram of the EMG data and stored histograms for locomotion steady states and locomotion transition states;

classifying using if-else rules and the matching scores the histogram as representing a locomotion steady state or locomotion transition state in the prosthetic device;

controlling the prosthetic device using the computed transitions between locomotion modes.