Screening for neurological disease using speech articulation characteristics

• US 10,010,288 B2
• Filed: 01/17/2017
• Issued: 07/03/2018
• Est. Priority Date: 10/16/2012
• Status: Active Grant

• Current Assignee: The Board of Trustees of Michigan State University (Michigan State University), University of Florida Research Foundation, Incorporated (State University System of Florida)
• Original Assignee: The Board of Trustees of Michigan State University (Michigan State University), University of Florida Research Foundation, Incorporated (State University System of Florida)
• Inventors: Rosenbek, John Clyde, Skowronski, Mark D., Shrivastav, Rahul, Anand, Supraja
• Primary Examiner(s): AZAD, ABUL K
• Application Number:
  US15/407,765
• Publication Number:
  US 20170119302A1
• Time in Patent Office: 532 Days
• CPC Class Codes: A61B 5/4082 Diagnosing or monitoring movement diseases, e.g. Parkinson, Huntington or Tourette
• CPC Class Codes: A61B 5/4803 Speech analysis specially adapted for diagnostic purposes
• CPC Class Codes: A61B 5/7275 Determining trends in physiological measurement data; Predicting development of a medical condition based on physiological measurements, e.g. determining a risk factor
• CPC Class Codes: A61B 7/00 Instruments for auscultation
• CPC Class Codes: A61B 7/04 Electric stethoscopes microphones, acoustic transducers therefor H04R

Abstract

Detection of neurological diseases such as Parkinson'"'"'s disease can be accomplished through analyzing a subject'"'"'s speech for acoustic measures based on human factor cepstral coefficients (HFCC). Upon receiving a speech sample from a subject, a signal analysis can be performed that includes identifying articulation range and articulation rate using HFCC and delta coefficients. A likelihood of Parkinson'"'"'s disease, for example, can be determined based upon the identified articulation range and articulation rate of the speech.

Claims

1. A method for screening for diseases, the method comprising:

performing a signal analysis of a speech sample received from a subject, the signal analysis comprising extracting cepstral coefficients from the speech sample and identifying articulation range and articulation rate using the cepstral coefficients extracted from the speech sample, analyzing language patterns, and combining the signal analysis of the speech sample and the analyzed language patterns by a coding module, wherein the analyzing of the language patterns comprises analyzing the language patterns to determine if an indicator of a neurological disease is present, and wherein the signal analysis of the speech sample and the analyzing of the language patterns are interconnected with each other;

determining a likelihood and type of a disease based upon the articulation range and articulation rate identified by the signal analysis of the speech sample, as well as the analyzing of the language patterns; and

outputting information indicating the likelihood and type of the disease,

wherein the cepstral coefficients from the speech sample are human factor cepstral coefficients or mel-frequency cepstral coefficients.

2. The method according to claim 1, wherein identifying the articulation range and articulation rate using the cepstral coefficients extracted from the speech sample comprises:

calculating standard deviation sums of the cepstral coefficients to estimate the articulation range; and

calculating delta coefficients of the speech sample to estimate the articulation rate.

3. The method according to claim 1, wherein extracting the cepstral coefficients from the speech sample comprises:

constructing a spectrogram of a signal of the speech sample;

calculating a log-power for each frame from the spectrogram;

applying a filter to a magnitude of the log-power;

performing a log-compression and transform to a cepstral domain of an output of the filter; and

forming a vector of cepstral coefficients for each frame.

4. The method according to claim 1, wherein the signal analysis of the speech sample does not involve performing boundary detection or phonetic analysis.

5. The method according to claim 1, wherein determining the likelihood and type of the disease based upon the articulation range and articulation rate identified by the signal analysis of the speech sample comprises using the articulation range and articulation rate to determine a likelihood of Parkinson'"'"'s Disease for the subject.

6. The method according to claim 5, wherein determining the likelihood of Parkinson'"'"'s disease comprises:

comparing the articulation range and articulation rate of the speech sample with normative baseline acoustic measures for Parkinson'"'"'s disease to generate values indicating the likelihood of Parkinson'"'"'s disease for the subject.

7. The method according to claim 1, wherein determining the likelihood and type of the disease based upon the articulation range and articulation rate identified by the signal analysis of the speech sample comprises:

comparing the articulation range and articulation rate of the speech sample with previously obtained measures from the same subject to generate values indicating the likelihood of Parkinson'"'"'s disease for the subject.

8. The method according to claim 1, wherein determining the likelihood and type of the disease based upon the articulation range and articulation rate identified by the signal analysis of the speech sample comprises:

comparing the articulation range and articulation rate of the speech sample with corresponding measures obtained from a set of other speakers to generate values indicating the likelihood of Parkinson'"'"'s disease for the subject.

9. The method according to claim 1, wherein the speech sample is received from an outgoing call of the subject received through a telephony, voice over internet protocol (VoIP) or cellular service provider.

10. The method according to claim 1, wherein the speech sample is received as a recorded speech sample uploaded to a network via a website interface.

11. The method according to claim 1, wherein the speech sample is received directly from a microphone.

12. The method according to claim 1, further comprising:

performing a pre-processing of the speech sample to clean the speech sample or select segments of the speech sample before performing the signal analysis.

13. The method according to claim 1, wherein the determination of a likelihood and type of a disease is based upon the articulation range, articulation rate, and articulation accuracy identified by the signal analysis of the speech sample.

14. A system for screening for a disease, the system comprising:

an application service provider for receiving a speech sample from a subject;

a memory for receiving and storing the speech sample; and

one or more computer-readable storage media in operable communication with the memory and having stored thereon computer-executable instructions comprising;

a pre-processing module for receiving the speech sample from the application service provider and cleaning the speech sample or selecting segments of the speech sample for further processing;

a speech metric module for receiving the speech sample from the pre-processing module and identifying articulation range and articulation rate using cepstral coefficients; and

a language marker module for receiving the speech sample from the pre-processing module and analyzing language patterns;

a modeling and coding module receiving an output of the speech metric module and an output of the language marker module;

a comparator for comparing the output of the speech metric module and the output of the language marker module with normative data, criteria, or previous output of the speech metric module and the language marker module stored in the memory of the system and outputting a decision indicating a likelihood of a neurological disease,

wherein the cepstral coefficients are human factor cepstral coefficients or mel-frequency cepstral coefficients, and

wherein the language marker module is interconnected with the speech metric module.

15. The system according to claim 14, wherein the speech metric module identifies the articulation range and articulation rate using the cepstral coefficients by:

calculating standard deviation sums of the cepstral coefficients to estimate the articulation range; and

calculating delta coefficients of the speech sample to estimate the articulation rate.

16. The system according to claim 14, wherein the speech metric module extracts the cepstral coefficients from the speech sample and uses the extracted cepstral coefficients to identify the articulation range and articulation rate.

17. The system according to claim 16, wherein the speech metric module extracts the cepstral coefficients from the speech sample by:

constructing a spectrogram of a signal of the speech sample;

calculating a log-power for each frame from the spectrogram;

applying a filter to a magnitude of the log-power;

performing a log-compression and transform to a cepstral domain of an output of the filter; and

forming a vector of cepstral coefficients for each frame.

18. The system according to claim 14, wherein the identifying articulation range and articulation rate using cepstral coefficients does not involve performing boundary detection or phonetic analysis.

19. The system according to claim 14, wherein the disease is Parkinson'"'"'s disease.

20. The system according to claim 14, wherein the speech metric module identifies articulation range, articulation rate, and articulation accuracy using the cepstral coefficients.