Wearable respiratory monitoring system based on resonant microphone array
A method for continuous acoustic signature recognition and classification includes a step of obtaining an audio input signal from a resonant microphone array positioned proximate to a target, the audio input signal having a plurality of channels. The target produces characterizing audio signals depending on a state or condition of the target. A plurality of features is extracted from the audio input signal with a signal processor. The plurality of features is classified to determine the state of the target. An acoustic monitoring system implementing the method is also provided.
1. A method for continuous acoustic signature recognition and classification, comprising:
obtaining an audio input signal from a resonant microphone array positioned proximate to a target, the audio input signal having a plurality of channels, the target producing characterizing audio signals depending on a state or condition of the target, wherein the resonant microphone array provides prefiltering of the audio input signals;
extracting a plurality of acoustic features from the audio input signal with a signal processor, wherein the acoustic features are characterized by a unique distribution of acoustic energy in a time domain waveform, in a spectral domain, or in variation in the spectral domain over time, wherein extracting acoustic features from the audio input signal comprises directly calculating a plurality of filter energies at a plurality of filter bank frequencies from the time domain waveform without discrete cosine transform or digital bandpass filtering steps;
classifying the plurality of acoustic features into acoustic classifications; and
continuously monitoring the state of the target from the acoustic classifications.
2. The method of claim 1 wherein the target is a human subject and the audio input signal is produced by respiration, the resonant microphone array being positioned proximate to a human subject's chest to provide for monitoring respiration of the human subject.
3. The method of claim 1 wherein the resonant microphone array includes a plurality of resonant microphones.
4. The method of claim 1 wherein each resonant microphone has a resonant frequency such that resonant frequencies of a plurality of resonant microphones are spaced between 20 Hz and 12,000 Hz.
5. The method of claim 1 wherein the resonant microphone array includes a plurality of capacitive combinations of diaphragm and back plate for resonant microphones.
6. The method of claim 1 wherein the audio input signal is pre-filtered by the resonant microphone array to produce a pre-filtered audio input signal.
7. The method of claim 6 wherein the step of extracting a plurality of features from the audio input signal comprises:
partitioning the pre-filtered audio input signal into one or more frames;
applying a windowing function to each of the one or more frames of the pre-filtered audio input signal;
performing pre-emphasis on the pre-filtered audio input signal that has been windowed and framed;
transforming the pre-filtered audio input signal into a feature vector; and
outputting the feature vector.
8. The method of claim 7 wherein transforming the pre-filtered audio input signal into the feature vector includes directly calculating a spectral energy of the pre-filtered audio input signal in the time domain.
9. The method of claim 7 wherein transforming the pre-filtered audio input signal into the feature vector includes applying an inverse discrete cosine transform and then calculating a vector of cepstral coefficients comprising the feature vector.
10. The method of claim 7 wherein a pre-emphasis network implements the step of performing pre-emphasis on the pre-filtered audio input signal.
11. The method of claim 7 wherein performing pre-emphasis on the pre-filtered audio input signal includes boosting one or more frequencies of the pre-filtered audio input signal relative to other frequencies of the pre-filtered audio input signal to minimize transmission distortion.
12. The method of claim 1 wherein classifying the plurality of acoustic features to determine the state of the target is performed by a classification algorithm.
13. The method of claim 12 wherein the classification algorithm is selected from the group consisting of Naive Bayes classifiers, Hidden-Markov Models, Artificial Neural Networks, and Support Vector Machines.
14. The method of claim 1 wherein the target is a fluid conduit.
15. The method of claim 1 wherein the step of monitoring the state of the target from the acoustic classifications includes a step of monitoring respiration in a mammalian subject.
16. The method of claim 15 wherein asthma and wheezing are monitored.
17. The method of claim 1 wherein the step of monitoring the state of the target from the acoustic classifications includes a step of automatically detecting medical acoustic signatures.
18. The method of claim 1 wherein the step of monitoring the state of the target from the acoustic classifications includes a step of monitoring heart sounds to detect murmurs.
19. An acoustic monitoring system, comprising:
a resonant microphone array configured to:
obtain an audio input signal having a plurality of channels, and
pre-filter the audio input signal through the resonant microphone array to form a pre-filtered audio input signal(s); and
a signal processor configured to:
divide the pre-filtered audio input signal into one or more frames,
windowing each of the one or more frames of the pre-filtered audio input signal,
transform the pre-filtered audio input signal into a feature vector that is outputted,
determine whether the audio input signal matches with a specific sound or sounds through a classification algorithm, wherein the feature vector includes acoustic features that are characterized by a unique distribution of acoustic energy in a time, domain waveform, in a spectral domain, or in variation in the spectral domain over time, wherein extracting acoustic features from the audio input signal comprises directly calculating a plurality of filter energies at a plurality of filter bank frequencies from the time domain waveform without discrete cosine transform or digital bandpass filtering steps, and
continuously monitor the state of a target by wirelessly sending the audio input signal or a notification of detection out to a nearby mobile phone or wireless transceiver.
20. The acoustic monitoring system of claim 19 wherein the classification algorithm is a machine-learning algorithm.
21. The acoustic monitoring system of claim 19 wherein the classification algorithm is selected from the group consisting of Naive Bayes classifiers, Hidden-Markov Models, Artificial Neural Networks, and Support Vector Machines.
22. The acoustic monitoring system of claim 19 wherein the resonant microphone array has a plurality of piezoelectric cantilevers or capacitive diaphragms or any other structures that form a plurality of resonant microphones that obtain audio signals and pre-filter the audio signals into a plurality of frequency bands.
23. The acoustic monitoring system of claim 22 , wherein each resonant microphone or piezoelectric cantilever has a resonant frequency, wherein the resonant frequencies of the plurality of resonant microphones or piezoelectric cantilevers are spaced between 20 Hz and 12,000 Hz.
24. The acoustic monitoring system of claim 19 wherein the resonant microphone array includes a capacitive combination of diaphragm and back plate.
25. The acoustic monitoring system of claim 19 further comprising an adhesive layer or strap for attaching the resonant microphone array to a human subject.