Signal processing method and abnormal sound detection system

An abnormal sound detection system, comprising:

• a sound receiving apparatus, which is disposed below a device under test for receiving a sound;
• and
• an electronic device, comprising:
• a receiving port, configured to receive a sound signal from the sound receiving apparatus;
• a storage device, storing a training database and a neural network model;
• and
• a processor, coupled to the receiving port and the storage device, and configured to:
• train the neural network model, wherein the neural network model comprises an input layer, a first hidden layer, a second hidden layer, and an output layer, and the first hidden layer, the second hidden layer, and the output layer have individual activation functions respectively;
• receive the sound signal;
• input the sound signal into the trained neural network model to output a probability value;
• determine whether the probability value is within one of a plurality of numerical ranges, wherein the plurality of numerical ranges correspond to a plurality of notification signals;
• and
• output a corresponding one of the notification signals based on one numerical range in which the probability value is located,
• wherein training the neural network model comprises:
• (a) randomly selecting a plurality of sample signals from the training database to obtain a combined signal, wherein the training database comprises a first sample set belonging to a first classification label and a second sample set belonging to a second classification label, and the number of the selected sample signals conforms to a preset number;
• (b) repeating the step (a) for a plurality of times to obtain a plurality of combined signals;
• and
• (c) using the combined signals to train the neural network model.