It is strongly recommended that a speech-recognition engine support the following waveform-audio formats:
· 11 kilohertz (kHz), 8-bit mono pulse code modulation (PCM). Every sound card can support this format.
· 11 kHz, 16-bit mono PCM. Most sound cards can support this waveform-audio format.
· 8 kHz, 8/16-bit mono PCM. Telephony cards can output this format.
An engine can also support the following waveform-audio formats:
· 16 kHz or 22 kHz, 8/16-bit mono PCM. The higher fidelity increases accuracy.
· Custom formats. Supporting one or more custom formats allows an engine to take advantage of digital signal processing (DSP) hardware that has waveform-audio drivers that support the custom formats.