Abstarct: The problem of acoustic detection and recognition is applied in robot audition, surveillance systems and security systems. This thesis addresses the problem of automatic sound detection and recognition of impulsive sounds. Unlike other recognition techniques, which extract features form signal’s frxame, our proposed system extracts features from input audio signals without framing them. In this thesis a novel feature extraction method with low level feature dimension is proposed. Our proposed system has low level computational load, there for it is suitable for online applications. The proposed system consists of detection and recognition stages. In detection stage system finds out whether a received sound is impulsive or not. For detection purpose we proposed two novel approaches baxsed on power evolution of input audio signal. Detection rate of our approaches is 100%. When detection algorithm finds an impulsive sound, recognition stage is triggered in order to classify incoming sound. Performance of our classification method, baxsed on signal’s behavior is evaluated using KNN classifier that accuracy of classification rate is achieved 93.75%. Our proposed classification method performs well even under noise condition. The developed system has a 93.75% and 71.87% classification rate at 50dB and 0dB white Gaussian noise degradation, respectively