Interference alignment (IA) consolidates interference signals into a small subspace and makes the interference-free detection possible for useful signals. Most existing IA precoders such as CJ-IA are designed to maximize the degree of freedom (DoF), which does not necessarily result in optimized system performance such as receiving signal-to-noise ratio (SNR). In this paper, based on CJ-IA, a maximum projection based IA (MP-IA) precoding scheme is proposed to maximize the desired signal strength at receivers, which brings optimized SNR, sum-rate and bit error rate (BER) performance. It is shown in simulations that the proposed MP-IA outperforms existing schemes in terms of desired signal strength, sum-rate and BER.