The active constellation extension (ACE) technique is a lossless (in terms of throughput) peak-to-average power ratio (PAR) reduction technique that adaptively extends the signal constellation while ensuring that the minimum distance between any two constellation points does not decrease. However, this technique increases the average transmit power. In this paper, we propose an adaptive-scaling algorithm for the implementation of ACE. This algorithm, based on the clipping and filtering technique, uses only the peak samples of clipping noise to reduce PAR. Simulation results show that the proposed algorithm has better PAR reduction, lower complexity and smaller BER than the previously proposed Smart Gradient-Project ACE algorithm.