Apparatus and method for diagnosing failure of microphone of vehicle
An apparatus and method for diagnosing a microphone of a vehicle reduce an interior noise by applying an active road surface noise reduction method, learn sounds collected using a plurality of microphones to generate a simulated signal for a specific microphone, and detect a failure of the specific microphone by comparing the simulated signal with an actually measured sound of the specific microphone so that noise reduction efficiency can be improved by distinguishing a noise from an interior sound, a failed microphone can be accurately detected using the sounds of the plurality of microphones, the spread of an unnecessary noise can be prevented by diagnosing a failure of the specific microphone, and a more comfortable and stable interior environment can be provided to occupants.
1 . An apparatus for diagnosing a microphone of a vehicle comprising (1) a plurality of microphones including a first microphone and (2) a plurality of speakers, the apparatus comprising:
a processor; and
a non-transitory computer-readable medium storing executable instructions which, when executed by the processor, cause the processor alone or in combination with other processors to control the apparatus to perform:
processing a sound captured by each of the microphones to generate a digital signal from each of the microphones to provide a plurality of digital signals representing the sound;
generating a simulated signal for the first microphone using a learning process based on the digital signals of the plurality of microphones excluding the digital signal of the first microphone;
comparing a measured signal of the first microphone with the simulated signal to calculate a loss value of the first microphone; and
determining, based on the calculated loss value, whether a failure has occurred to the first microphone.
2 . The apparatus of claim 1 , wherein, for detecting whether a failure has occurred to the first microphone, the instructions, when executed by the processor, further cause the processor alone or in combination with other processors to control the apparatus to perform:
determining whether the calculated loss value of the first microphone exceeds a preset threshold value; and
in response to determining that the calculated loss value of the first microphone exceeds the preset threshold value, diagnosing that the failure has occurred to the first microphone.
3 . The apparatus of claim 2 , wherein the instructions, when executed by the processor, further cause the processor alone or in combination with other processors to control the apparatus to perform:
comparing the simulated signal for the first microphone with the measured signal of the first microphone through a loss function to calculate the loss value of the first microphone;
optimizing, via a learning process, the loss value to calculate a plurality of loss values;
adding a predetermined margin to a maximum value among the plurality of loss values; and
setting, based on the predetermined margin added to the maximum value, the threshold value for the first microphone.
4 . The apparatus of claim 3 , wherein, for generating the simulated signal for the first microphone, the instructions, when executed by the processor, further cause the processor alone or in combination with other processors to control the apparatus to perform generating, using a long short-term memory (LSTM) auto encoder, the simulated signal for the first microphone.
5 . The apparatus of claim 4 , wherein:
the LSTM auto encoder includes a plurality of LSTM encoders and a plurality of LSTM decoders,
each LSTM encoder is configured to extract and abstract a feature of a signal, and
each LSTM decoder is configured to restore and actualize the signal.
6 . The apparatus of claim 4 , wherein, for optimizing the loss value, the instructions, when executed by the processor, further cause the processor alone or in combination with other processors to control the apparatus to perform feeding back a network weight to the LSTM auto encoder using stochastic gradient descent based on the loss value.
7 . The apparatus of claim 1 , wherein the instructions, when executed by the processor, further cause the processor alone or in combination with other processors to control the apparatus to perform:
detecting whether a failure has occurred to the plurality of microphones; and
in response to detecting that a failure has occurred to at least one of the plurality of microphones, stopping a noise reduction operation and outputting a warning for the microphone to which the failure has occurred.
8 . The apparatus of claim 7 , wherein the instructions, when executed by the processor, further cause the processor alone or in combination with other processors to control the apparatus to perform:
analyzing the sounds captured by the plurality of microphones to classify a type and amplitude of each captured sound;
detecting a road surface noise to generate an attenuation signal with a phase opposite to a phase of the road surface noise; and
outputting the attenuation signal through the plurality of speakers to perform the noise reduction operation.
9 . A method of diagnosing a microphone of a vehicle, the vehicle comprising (1) a plurality of microphones including a first microphone and (2) a plurality of speakers, the method comprising:
processing a sound captured by each of the microphones to generate a digital signal from each of the microphones to provide a plurality of digital signals representing the sound;
generating a simulated signal for the first microphone using a learning process based on the digital signals of the plurality of microphones excluding the digital signal of the first microphone;
comparing a measured signal of the first microphone with the simulated signal to calculate a loss value of the first microphone; and
determining, based on the calculated loss value, whether a failure has occurred to the first microphone.
10 . The method of claim 9 , wherein determining whether the failure has occurred to the first microphone comprises:
determining whether the loss value of the first microphone exceeds a preset threshold value; and
in response to determining that the calculated loss value exceeds the preset threshold value, diagnosing that the failure has occurred to the first microphone.
11 . The method of claim 9 , wherein:
the plurality of microphones includes second, third and fourth microphones, and
generating the simulated signal comprises generating the simulated signal based on signals of the second, third and fourth microphones.
12 . The method of claim 11 , wherein generating the simulated signal includes generating the simulated signal of the first microphone based on the signals of the second, third and fourth microphones using a long short-term memory (LSTM) auto encoder.
13 . The method of claim 12 , wherein:
the LSTM auto encoder includes a plurality of LSTM encoders and a plurality of LSTM decoders, and
generating the simulated signal includes:
extracting and abstracting features of the signals using the plurality of LSTM encoders; and
restoring and actualizing the signals using the plurality of LSTM decoders.
14 . The method of claim 12 , further comprising:
comparing the simulated signal of the first microphone with the measured signal of the first microphone through a loss function to calculate the loss value of the first microphone;
optimizing the loss value by feeding back a network weight to the LSTM auto encoder using stochastic gradient descent based on the loss value;
performing learning by repeating generating the simulated signal to the optimizing of the loss value; and
setting a threshold value for the first microphone by adding a predetermined margin to a maximum value among a plurality of calculated loss values.
15 . The method of claim 9 , further comprising:
detecting whether a failure has occurred to the plurality of microphones; and
in response to detecting that a failure has occurred to at least one of the plurality of microphones, performing (1) stopping a noise reduction operation, and (2) outputting a warning for the microphone to which the failure has occurred.
16 . The apparatus of claim 1 , wherein, for generating the simulated signal for the first microphone, the instructions, when executed by the processor, further cause the processor alone or in combination with other processors to control the apparatus to perform generating, using a long short-term memory (LSTM) auto encoder, the simulated signal for the first microphone.
17 . The method of claim 9 , wherein generating the simulated signal includes generating the simulated signal of the first microphone based on the signals of the plurality of microphones excluding the first microphone using a long short-term memory (LSTM) auto encoder.
18 . The apparatus of claim 1 , wherein the learning process is performed at a manufacturing stage of the vehicle.
19 . The apparatus of claim 18 , wherein learning data for the learning process is updated by repeating the learning process based on data collected during traveling of the vehicle.
20 . The method of claim 9 , wherein the learning process is performed at a manufacturing stage of the vehicle.