The dominant technological approaches for speech recognition system are based on pattern matching of statistical representations of the acoustic speech signal. However, although significant progress has been made in the field of ASR these last years, the typical vocabulary size is still very limited and the performance of the resulting systems is still not comparable to that achieved by human beings[15]. Here, we describes what kinds of speech recognition methods are adopted in the CAPL system in order to avoid the unreliable aspects of current ASR. The structure of speech recognition methods for the CAPL system is illustrated in Figure 2.4.
図 2.4: Structure of speech recognition methods for the CAPL system