Diagnostic techniques based on speech models

A method, comprising:

• obtaining at least one speech model that:
• includes one or more acoustic states exhibited in one or more reference speech samples, the acoustic states being associated with respective local distance functions such that, given any acoustic feature vector within a domain of the local distance functions, the local distance function of each acoustic state returns a local distance indicating a degree of correspondence between the given acoustic feature vector and the acoustic state, and
• defines allowed transitions between the acoustic states, provided that the speech model includes multiple acoustic states;
• receiving at least one test speech sample produced by a subject;
• computing a plurality of test-sample feature vectors that quantify acoustic features of different respective portions of the test speech sample;
• based on the local distance functions and on the allowed transitions, mapping the test speech sample to a minimum-distance sequence of the acoustic states, by mapping the test-sample feature vectors to respective ones of the acoustic states such that a first total distance between the test-sample feature vectors and the respective ones of the acoustic states is minimized,
• the first total distance being based on respective local distances between the test-sample feature vectors and the respective ones of the acoustic states;
• computing a second total distance between the test-sample feature vectors and the respective ones of the acoustic states, the second total distance being different from the first total distance;
• and
• responsively to the second total distance, generating an output indicating a physiological state of the subject.

A computer software product comprising a tangible non-transitory computer-readable medium in which program instructions are stored, which instructions, when read by a processor, cause the processor to:

• obtain at least one speech model that:
• includes one or more acoustic states exhibited in one or more reference speech samples, the acoustic states being associated with respective local distance functions such that, given any acoustic feature vector within a domain of the local distance functions, the local distance function of each acoustic state returns a local distance indicating a degree of correspondence between the given acoustic feature vector and the acoustic state, and
• defines allowed transitions between the acoustic states, provided that the speech model includes multiple acoustic states,
• receive at least one test speech sample produced by a subject,
• compute a plurality of test-sample feature vectors that quantify acoustic features of different respective portions of the test speech sample,
• based on the local distance functions and on the allowed transitions, map the test speech sample to a minimum-distance sequence of the acoustic states, by mapping the test-sample feature vectors to respective ones of the acoustic states such that a first total distance between the test-sample feature vectors and the respective ones of the acoustic states is minimized,
• the first total distance being based on respective local distances between the test-sample feature vectors and the respective ones of the acoustic states,
• compute a second total distance between the test-sample feature vectors and the respective ones of the acoustic states, the second total distance being different from the first total distance, and
• responsively to the second total distance, generate an output indicating a physiological state of the subject.