Fire Detection Algorithms Using Multimodal ... - Bilkent University
Fire Detection Algorithms Using Multimodal ... - Bilkent University
Fire Detection Algorithms Using Multimodal ... - Bilkent University
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
CHAPTER 5. FLAME DETECTION USING PIR SENSORS 75An HMM based classification is carried out for fire detection. Two three-stateMarkov models are used to represent fire and non-fire events (cf. Fig. 6.2). Inthese Markov models, state S1 corresponds to no activity within the viewing rangeof the PIR sensor. The system remains in state S1 as long as there is not anysignificant activity, which means that the absolute value of the current waveletcoefficient, |w[k]|, is below a non-negative threshold T 1. A second threshold T 2is also defined in wavelet domain which determines the state transitions betweenS2 and S3. If T 1 < |w[k]| < T 2, then state S2 is attained. In case of |w[k]| > T 2,state S3 is acquired.Figure 5.5: Two three-state Markov models are used to represent (a) ‘fire’ and(b) ‘non-fire’ classes, respectively.The first step of the HMM based analysis consists of dividing the waveletcoefficient sequences in windows of 25 samples. For each window, a correspondingstate transition sequence is determined. An example state transition sequence ofsize 5 may look likeC = (S2, S1, S3, S2, S1) (5.3)Since the wavelet signal captures the high frequency information in the signal,we expect that there will be more transitions occurring between states when