This paper proposes a new method for continuous acoustic adaptive feedback cancelation (AFC) in digital hearing aids. The proposed method employs two adaptive filters working in tandem. The first adaptive filter is excited by the receiver (output) signal of the hearing aid, and uses microphone signal as its desired response. The lattice-predictor based adaptive algorithm is used to updated the coefficients of the first adaptive filter. The second adaptive filter is excited by a (random) probe signal. We propose coefficient monitoring strategy with two fold objectives: 1) both adaptive filters converge to a good estimate of the acoustic feedback path, and that 2) both adaptive filter are re-initialized when a sudden change in the acoustic feedback path is detected. Finally, the injected probe noise is controlled via time-varying gain in such a way that a low level noise is used when the system is operating in its steady state. Simulation results demonstrate that the proposed method achieves good modeling accuracy, preserves good speech quality, and provides robust performance for the sudden changes in acoustic feedback path.