VSE knjižnice (vzajemna bibliografsko-kataložna baza podatkov COBIB.SI)
  • Identification of machinery sounds [Elektronski vir]
    Lipar, Primož ...
    Speaker recognition methods are well knownand widely used in the ASR (Automatic Speech Recognition)systems. The use of these methods for the classification ofmachinery sounds in noisy environments is ... presented in thispaper. Influence of background noise was reduced by using ahighly directive sound recording, which can be understood asa spatial filter. A fusion of microphone antenna withbeamforming algorithm forms such a filter, which improvesSNR (Signal to Noise Ratio). Features of machinery soundshave been extracted using standard MFCC (Mel FrequencyCepstral Coefficients) parameterization method with Meland linear frequency scaling. SVM (Support VectorMachine) classifier was used for the classification of soundfeatures. A significant improvement of the classifier decisionperformance was achieved in noisy environment when 8microphones were used together with beamformingalgorithm. Results of using Mel and linear scale are alsopresented and show similar results in recognition ofmachinery sounds.
    Vir: TELFOR 2013 [Elektronski vir] : proceedings of papers (Datoteka TELFOR2013_05_17 (4 f.))
    Vrsta gradiva - prispevek na konferenci
    Leto - 2013
    Jezik - angleški
    COBISS.SI-ID - 13271323

Vnos na polico