Computer recognition of human speech