Measurement noise intensity is generally high and irregular in the ocean when a passive sonar system is used to track targets. Especially, the characteristics of measurement noises are not known accurately. An existing TMA method uses a sequential estimator called the modified gain extended Kalman filter (MGEKF) with measurement noise variance predetermined from operational test. If the predetermined measurement noise variance are mismatched to the real world, TMA may have degraded. In order to identify the measurement noise variance adaptively, the interacting multiple model (IMM) using 2 MGEKFs is suggested. The proposed TMA method is tested by a series of simulation runs.