Speech quality of VoIP is degraded by transmission errors such as packet loss that is inevitable in best-effort communications. We have studied how to decrease such degradation. This study investigated PWR (Pitch Waveform Replication) method that is employed as an error hiding technique. The previous version of PWR employs an interpolation where past speech data is only used. For the decrease of some artifacts such as echoic quality in the begging and end of phonation, this study has proposed a new version that takes account of an interpolation from future speech data. In order to shorten the delay of the pitch extraction from future speech data, the proposed method employs a template matching. From the experimental results of objective and subjective evaluation, it is indicated that the proposed method is potentially useful for the improvement of speech quality.