The Normalized Quantized Gradient-Based adaptive algorithm(NQLMP)has many drawbacks. For example, it has high computational complexity and the statistical properties are poor, if the power p is high. In this paper the modification of the NQLMP algorithm(MNQLMP)is presented. The proposed algorithm uses the recursively first-order gradient signal energy to control the coefficient updating instead the recursively p-order gradient signal energy. This method caused the proposed algorithm has higher convergence speed and high accuracy of the filter coefficient more than the NQLMP algorithm. The results of computer simulation and Real-Time implementation are consistency which attest the MNQLMP algorithm is better than the NQLMP algorithm.