Sensors based on the magnetoelectric (ME) effect have the potential to be genuine alternatives for measuring bio-magnetic signals. Unfortunately, the sensor structure usually inhibits the problem that several non-magnetic types of noise couple mechanically into the sensor: in this contribution, we will focus on undesired acoustic coupling. Therefore, an adaptive cancellation approach based on a computationally efficient gradient estimation algorithm with a pseudo-optimally control scheme is proposed. The approach is using a microphone as a noise reference sensor and is implemented in real time. An evaluation in terms of measurements is performed inside a magnetically shielded chamber. For a particular scenario, which is characterized by double excitation, an algorithm with binary control-scheme improves the signal-to-noise ratio (SNR) only by around 4dB. If the proposed control scheme is used instead, an improvement of the SNR of around 13dB is achieved.