Evaluating the detection performance of knocking sounds! Introducing deep learning that has achieved a high abnormality detection rate.
The knocking sound of the engine can be detected using acceleration sensors and combustion pressure sensors, but in places like engine durability testing, detection using only non-contact microphones is expected.
Here, the application of machine learning is considered, but it is difficult to collect a large amount of knocking sound as training data.
Therefore, we prepared the sound of striking metal cans, created a framework to extract the characteristics of impact sounds, applied this framework to knocking sounds, and achieved a recognition accuracy of about 80% for engine knocking sounds.
【Features】
■ Evaluation function: ReLU function
■ Pooling function: Max pooling function
■ Discrimination: Softmax function
■ Fully connected layer: Dropout
*For more details, please refer to the PDF document or feel free to contact us.
- Company:キャテック
- Price:Other