FRS: Adaptive Score for Improving Acoustic Source Classification from Noisy Signals

This paper proposes a frame relevance score to improve the classification of environmental acoustic sources from noisy speech signals. The importance of each short-time frame for the classification results is objectively interpreted by SHAP values. The SHAP-based frame relevance score enables the selection of frames that are more appropriate to improve the discrimination power of the acoustic models. The frame selection can be used as a pre-training strategy to any classification strategy. Evaluation experiments consider the recognition of ten background sources from noisy speech signals. The classical approach based on MFCC and GMM is adopted to prove that the selected frames can better distinguish the acoustic classes. Moreover, the frame selection outperforms a surrogate-based adaptive learning solution. Experiments are also conducted with a recently proposed pre-trained neural network that achieves high classification rates. The proposed SHAP-based selection shows improved classification accuracies even for this scenario.