2014 : Spike Reduction using Expectation Maximization based on Kalman Filter

Dr. Ir. Yoyon Kusnendar Suprapto MSc.


Abstract

In the recording of gending-java is somewhat different with the international musical instruments, as gamelan instruments consisting mainly of metallophone instruments, sometimes occurring spike signal, in some circumstances the spike is too high, it will result in defects in the gending sound, in this research tried to reducing the value spike in the signal, so that its value approaches the value of the other signals. first step is modeling the signal with some spike signal, spike signals are further estimate by using expectation maximization (EM) based on kalman filter, the output signal is tested using the cosine similarity test, seen from the test results the average value is 0.96 close to 1, or similar to the original signal, so that EM kalman can be used to reducing the spike signal.