针对数字脉冲间隔调制(DPIM)包序列发生差错时会出现符号增减从而导致差错性能下降的问题，提出了一种新的解调方法。给出了解调方法的解调过程，推导了其差错解概率，并与门限解调法进行了比较分析。数值仿真表明，当误包率为10-6以及包信息为24 bit时，相较于门限法，可节省约1.5 dBm的平均接收功率，且调制阶数对其影响不大；但当包信息量增大时，性能改善的程度将减弱。在不增加冗余信息的情况下，提出的DPIM解调方法能够有效地改善系统的差错性能，可为无线光通信链路的设计提供参考。

There is a fluctuation of symbols which results in error performance degradation if some errors occur in digital pulse interval modulation (DPIM) train. To address this issue, we present a novel demodulation scheme for DPIM. The decoding method is firstly proposed, and then the packet error rate is derived and compared with decoding using the optimal threshold method. Numerical simulation shows that the proposed method saves approximately 1.5 dBm average signal power than that of threshold method with different modulation levels when the packet error rate is 10-6 and information of packet n is 24 bit, but the improvement of error performance will degrade with the increase of n. Requiring no redundant information bits, the proposed method can effectively improve the error performance for DPIM, and can help to design wireless optical communication systems on modulation.