The key of MBE algorithm is the continuous property of synthesis speech. This papaer presentsa new algorthm based on MBE. It propses using narrow-band random signal to produce unvoiced speech and anincreasing-decreasing factor to control voiced and unvoiced speech synthesis so that voiced and unvoiced speechsynthesis can be implemented on time field. This new algorithm improves the continuous properties of pitches andamplitudes of synthesis speech. The informal listening at 4. 8Kbps shows that the synthesis speech of our new al gorithm is better than that of traditional MBE.