Abstract: A method and apparatus for conducting point-of-sale transactions using both speaker dependent and speaker independent voice recognition in which a spoken utterance from a first user is captured with a sound input device. The spoken utterance is compared to a first plurality of stored patterns to find a first match. A product order is generated based on the first match found in the comparing step. The comparing and generating steps may be repeated for each additional product purchased. A bill is then calculated based on the product orders. The plurality of stored patterns, also referred to as a voice layer, comprises a subset of a user vocabulary stored in a memory. The voice layer, i.e., the plurality of stored patterns to which the spoken utterances are compared, changes to include different sets of likely words from the vocabulary depending upon the context of the transaction. In the speaker dependent embodiment, the user vocabulary is specific to the first user.