基于采用权电流D/A方式实现合成语音输出方式有集成电阻离散性大、开关的非线性，以及功耗高等缺点，提出了利用脉冲宽度调制(PWM)技术的方法将数字语音信号直接转换为脉冲宽度调制波，并通过低通滤波器生成模拟语音信号。PWM语音输出方式相对一般的DAC语音输出方式具有功耗小、工艺适应能力强、设计可靠等优点，并且PWM语音输出方式已经成功应用于语音合成芯片中，输出合成语音质量良好For the output mode of speech synthesis which has been come true by using traditional D/A method, there is a lot of disadvantages, such as high discrete integrate resistance, nonlinear switching, high power dissipation. We propose a new method based on the theory of pulse width modulation (PWM), which convert directly the digital speech signal to the pulse width modulation wave and become analog speech signal by passing low-pass filter. Comparing with the traditional D/A method, the new method has the advantages of low power dissipation, high technique adaptability and high designing reliability. It has been applied in the chips of speech synthesis successfully.