Share via


SPINTERFERENCE (SAPI 5.4)

Microsoft Speech API 5.4

SPINTERFERENCE

SPINTERFERENCE lists factors that can interfere with accurate recognition of speech input.

  
    typedef enum SPINTERFERENCE
{
    SPINTERFERENCE_NONE,
    SPINTERFERENCE_NOISE,
    SPINTERFERENCE_NOSIGNAL,
    SPINTERFERENCE_TOOLOUD,
    SPINTERFERENCE_TOOQUIET,
    SPINTERFERENCE_TOOFAST,
    SPINTERFERENCE_TOOSLOW,
    SPINTERFERENCE_LATENCY_WARNING,
    SPINTERFERENCE_LATENCY_TRUNCATE_BEGIN,
    SPINTERFERENCE_LATENCY_TRUNCATE_END
} SPINTERFERENCE;

Elements

  • SPINTERFERENCE_NONE
    Private event. Do not use.
  • SPINTERFERENCE_NOISE
    The sound received is interpreted by the speech recognition engine as noise. This event is generated when there is a SOUND_START followed by a SOUND_END without an intervening PHRASE_START. The event will be also generated during dictation if, after a series of hypotheses, it is determined that the signal is noise.
  • SPINTERFERENCE_NOSIGNAL
    A sound is received but it is of a constant intensity. This also includes the microphone being unplugged or muted.
  • SPINTERFERENCE_TOOLOUD
    A sound is received but the stream intensity is too high for discrete recognition.
  • SPINTERFERENCE_TOOQUIET
    A sound is received but the stream intensity is too low for discrete recognition.
  • SPINTERFERENCE_TOOFAST
    The words are spoken too quickly for discrete recognition.
  • SPINTERFERENCE_TOOSLOW
    The words are spoken too slowly and indicates excessive time between words.
  • SPINTERFERENCE_LATENCY_WARNING
    The duration of captured audio input that is still awaiting recognition has exceeded an initial threshold. SAPI takes no action, but client applications should be aware (and consumers should be made aware) that recognition will be delayed while processing catches up. Users should be informed that providing additional input may result in loss of input data (via a future truncation).
  • SPINTERFERENCE_LATENCY_TRUNCATE_BEGIN
    The duration of captured audio input that is still awaiting recognition has exceeded a higher threshold. SAPI discards subsequent input while it catches up on the data already in the audio queue. Users should be informed via the event that SAPI is not accepting audio input (new input will be lost) while it processes data that is already queued.
  • SPINTERFERENCE_LATENCY_TRUNCATE_END
    After SAPI begins truncating audio input (and raises a _LATENCY_TRUNCATE_BEGIN SPEI_INTERFERENCE event, if applicable), processing will catch up and the queue will be emptied. When no further data remains in the queue, SAPI will resume and raise _LATENCY_TRUNCATE_END to inform users and consumer applications that speech recognition will accept new audio input.

Remarks

Any of the SPINTERFERENCE elements may be the lParam of a SPEI_INTERFERENCE event.