Systems and methods for identifying patient distress based on a sound signal
A sound signal from a patient may include information that may be used to determine multiple patient parameters. A patient monitor may determine respiration information such as respiration rate from the sound signal, for example based on modulations of the sound signal due to patient breathing. The patient monitor may also determine indications of patient distress based on a trained classifier, speech commands, or sound patterns.
1. A method comprising:
receiving a sound signal from a sensor that senses sound from a patient, the sound signal indicative of human non-speech vocalization;
computing, with processing equipment, one or more metrics based on the sound signal;
determining, with the processing equipment, a classification of the sound signal based on the one or more metrics and on a classifier, wherein the classifier is trained based on signal characteristics that correspond to patient distress;
determining, with the processing equipment, whether the sound signal corresponds to patient distress based on the classification; and
outputting an indication of patient distress when patient distress is determined to be present.
2. The method of claim 1 , further comprising:
recognizing speech based on the sound signal;
determining a command based on the recognized speech; and
further determining whether the sound signal corresponds to patient distress based on the command.
3. The method of claim 2 , wherein the command comprises one or more of a request for assistance, an indication of pain level, and a request for medication.
4. The method of claim 1 , further comprising:
identifying a candidate portion of the sound signal; and
further determining whether the sound signal corresponds to patient distress based on the candidate portion.
5. The method of claim 4 , wherein the candidate portion is identified based on one or more of a sound level, pitch, frequency, a rate of change of the sound level, a rate of change of the pitch, and a rate of change of the frequency.
6. The method of claim 1 , wherein the classifier comprises one or more of a neural network, a genetic algorithm, stochastic classifiers, probabilistic classifiers, propositional logics, predicate logics, and nearest neighbor classification methods.
7. The method of claim 1 , further comprising:
processing the sound signal to generate a respiration signal; and
determining respiration information based on the respiration signal.
8. A non-transitory computer-readable storage medium for processing a sound signal, the computer-readable medium having computer program instructions recorded thereon for:
receiving a sound signal from a sensor that senses sound from a patient, the sound signal indicative of human non-speech vocalization;
computing one or more metrics based on the sound signal;
determining a classification of the sound signal based on the one or more metrics and on a classifier, wherein the classifier is trained based on signal characteristics that correspond to patient distress;
determining whether the sound signal corresponds to patient distress based on the classification; and
outputting an indication of patient distress when patient distress is determined to be present.
9. The computer-readable medium of claim 8 , the computer-readable medium having computer program instructions recorded thereon for:
recognizing speech based on the sound signal;
determining a command based on the recognized speech; and
further determining whether the sound signal corresponds to patient distress based on the command.
10. The computer-readable medium of claim 9 , wherein the command comprises one or more of a request for assistance, an indication of pain level, and a request for medication.
11. The computer-readable medium of claim 8 , the computer-readable medium having computer program instructions recorded thereon for:
identifying a candidate portion of the sound signal; and
further determining whether the sound signal corresponds to patient distress based on the candidate portion.
12. The computer-readable medium of claim 11 , wherein the candidate portion is identified based on one or more of a sound level, pitch, frequency, a rate of change of the sound level, a rate of change of the pitch, and a rate of change of the frequency.
13. The computer-readable medium of claim 8 , wherein the classifier comprises one or more of a neural network, a genetic algorithm, stochastic classifiers, probabilistic classifiers, propositional logics, predicate logics, and nearest neighbor classification methods.
14. A monitoring unit comprises processing equipment configured to:
receive a sound signal from a sensor that senses sound from a patient, the sound signal indicative of human non-speech vocalization;
compute one or more metrics based on the sound signal;
determine a classification of the sound signal based on the one or more metrics and on a classifier, wherein the classifier is trained based on signal characteristics that correspond to patient distress;
determine whether the sound signal corresponds to patient distress based on the classification; and
output an indication of patient distress when patient distress is determined to be present.
15. The monitoring unit of claim 14 , wherein the monitoring unit is further configured to:
recognize speech based on the sound signal;
determine a command based on the recognized speech; and
further determine whether the sound signal corresponds to patient distress based on the command.
16. The monitoring unit of claim 15 , wherein the command comprises one or more of a request for assistance, an indication of pain level, and a request for medication.
17. The monitoring unit of claim 14 , wherein the monitoring unit is further configured to:
identify a candidate portion of the sound signal; and
further determine whether the sound signal corresponds to patient distress based on the candidate portion.
18. The monitoring unit of claim 17 , wherein the candidate portion is identified based on one or more of a sound level, pitch, frequency, a rate of change of the sound level, a rate of change of the pitch, and a rate of change of the frequency.
19. The monitoring unit of claim 14 , wherein the monitoring unit is further configured to:
process the sound signal to generate a respiration signal;
determine respiration information based on the respiration signal.
20. The monitoring unit of claim 14 , wherein the classifier comprises one or more of a neural network, a genetic algorithm, stochastic classifiers, probabilistic classifiers, propositional logics, predicate logics, and nearest neighbor classification methods.