A new pruned-trellis search algorithm for high-rate convolutional code is developed. The search time and memory size is significantly reduced from standard search techniques. Some new high-rate systematic optimum convolutional codes of rate up to 7/8 have been found by this new search technique, and with constraint length up to 15. These newly found high-rate convolutional codes can be efficiently decoded using pruned, error-trellis, syndrome decoding. The real advantage of the pruned error-trellis decoding over the conventional Viterbi decoding algorithm is the reduction of the memory size required. Simulation shows that the error trellis performance of pruned error-trellis decoding suffers only a 0.2 dB loss for some systematic high-rate convolutional codes compared with conventional, full trellis decoding. Keywords: Integrated circuits; Architectures; Bibliographics; Abstracts.