Microphone defect detection
View Patent ↗Disclosed are an apparatus and method for testing a microphone of a device under test. The apparatus includes a speaker, a mount or bracket that secures a device under test a fixed distance from the speaker, and a controller. The controller causes the speaker to generate a test signal at various amplitudes. The controller receives an output signal generated by the microphone in response to the test signal and generates a prediction of whether the microphone is defective based on the output signal. The prediction may be generated by a machine-learning model such as a neural network or other trained classifier.
1. A method of testing a microphone, the method comprising:
generating a test signal at a constant frequency across a plurality of predetermined amplitudes, the plurality of predetermined amplitudes ranging from a first threshold amplitude to a second threshold amplitude;
receiving an output signal generated by the microphone in response to the test signal;
extracting a plurality of features from the output signal, the extracting comprising calculating a total harmonic distortion profile including a plurality of total harmonic distortion values, each total harmonic distortion value corresponding to one of the plurality of predetermined amplitudes, wherein the plurality of features include the plurality of total harmonic distortion values;
generating a feature vector of the plurality of features;
providing the feature vector of the plurality of features as input to a defect prediction model; and
receiving, from the defect prediction model, a prediction of whether the microphone is defective.
2. The method of claim 1 , wherein the test signal is a sine wave.
3. The method of claim 2 , wherein the frequency of the test signal is in a range from 500 hertz to 2 kilohertz.
4. The method of claim 1 , wherein the plurality of predetermined amplitudes increase in steps of equal amplitude from the first threshold amplitude to the second threshold amplitude.
5. The method of claim 1 , wherein the defect prediction model is a trained supervised learning model.
6. A microphone testing apparatus comprising:
a speaker;
a mount or bracket configured to secure a device under test a fixed distance from the speaker; and
a controller configured to:
cause the speaker to generate a test signal at a constant frequency across a plurality of predetermined amplitudes, the plurality of predetermined amplitudes ranging from a first threshold amplitude to a second threshold amplitude;
receive an output signal generated by a microphone of the device under test in response to the test signal; and
generate a prediction of whether the microphone is defective based on the output signal by:
extracting a plurality of features from the output signal, the extracting comprising calculating a total harmonic distortion profile including a plurality of total harmonic distortion values, each total harmonic distortion value corresponding to one of the plurality of predetermined amplitudes, wherein the plurality of features include the plurality of total harmonic distortion values,
generating a feature vector of the plurality of features, and
providing the feature vector of the plurality of features as input to a defect prediction model, the defect prediction model outputting the prediction of whether the microphone is defective.
7. The microphone testing apparatus of claim 6 , wherein the speaker and the mount or bracket are within an anechoic box.
8. The microphone testing apparatus of claim 6 , wherein the fixed distance is in a range from one centimeter to fifty centimeters.
9. The microphone testing apparatus of claim 6 , wherein the controller is software executing on the device under test.
10. The microphone testing apparatus of claim 6 , wherein the test signal is a sine wave of fixed frequency in a range from 500 hertz to 2 kilohertz.
11. The microphone testing apparatus of claim 6 , wherein the defect prediction model is a clustering model.
12. The microphone testing apparatus of claim 6 , wherein the plurality of predetermined amplitudes increase in steps of equal amplitude from the first threshold amplitude to the second threshold amplitude.
13. A non-transitory, computer-readable medium storing instructions that, when executed by a computing device, cause the computing device to perform operations comprising:
generating a test signal at a plurality of amplitudes at a constant frequency across a plurality of predetermined amplitudes, the plurality of predetermined amplitudes ranging from a first threshold amplitude to a second threshold amplitude;
receiving an output signal generated by the microphone in response to the test signal;
extracting a plurality of features from the output signal, the extracting comprising calculating a total harmonic distortion profile including a plurality of total harmonic distortion values, each total harmonic distortion value corresponding to one of the plurality of predetermined amplitudes, wherein the plurality of features include the plurality of total harmonic distortion values;
generating a feature vector of the plurality of features;
providing the feature vector of the plurality of features as input to a defect prediction model; and
receiving, from the defect prediction model, a prediction of whether the microphone is defective.
14. The non-transitory, computer-readable medium of claim 13 , wherein the plurality of predetermined amplitudes increase in steps of equal amplitude from the first threshold amplitude to the second threshold amplitude.
15. The non-transitory, computer-readable medium of claim 13 , wherein the test signal is a sine wave of fixed frequency in a range from 500 hertz to 2 kilohertz.
16. The non-transitory, computer-readable medium of claim 13 , wherein the defect prediction model is a trained supervised learning model.