Time-frequency clustering for burst gravitational waves search in TAMA300 data

We have developed a method 'time-frequency (TF) clustering' to find the burst gravitational waves for TAMA data analysis. TF clustering method on sonogram (spectrogram) shows some characteristics of short duration signal. Using parameters which represent the cluster shape, we can efficiently identify some predicted gravitational wave forms and can exclude typical unstable spike like noises due to detector instruments. The requirement of some parameters of cluster achieved roughly 50% average efficiency for injected DFM waveforms of $h_{rss}\sim 2 \times 10^{-19}$ for type I burst. Also the reduction for signal by spike noises are more than one order improvement for the SNR$>$100.