FIELD: physics.
SUBSTANCE: following steps are executed in order to classify sound-generating processes (P): detecting (S1) sound signals arising from sound-generating processes (H), extracting (S2) features (m) from the detected signals for the corresponding sound-generating process (P), calculating (S3) similarity between the detected sound signals based on the extracted features (m) in order to group processes (P) with similar sound signals into process clusters (C), adapting (S4) statistical distribution parameters of the reference model depending on statistical distribution parameters of those processes (P) contained in a process cluster (C) with a maximum number (Z) of processes (P) grouped therein in order to generate an adapted model, and classifying (S5) sound-generating processes (P) based on the adapted model.
EFFECT: enabling classification of sound-generating processes with automatic model adaptation with negligible computational costs, while guaranteeing error-free classification of sound-generating processes also with rapidly varying ambient conditions.
20 cl, 10 dwg
