| 摘要 |
An answer supervision method which has the steps of (1) Initialize and test all RAM and programmable hardware, (2) Establish software threshold for each channel and store in a Status Control Block, (3) Begin state driven analysis of cadences, (4) Analyze cadences without reporting results (Reset C), (5) Upon Reset A, try to detect specified duration of "heard" signal, (6) Immediately following detection, verify specified duration of silence, (7) Upon Reset B, sample PCM and determine if tone or silence is present, using software threshold. Save history of samples (in all cases of detection) and continuously analyze in order to detect tone cadences or presence of voice, send report to SBD when analysis is complete, (8) Return to Reset C state. Analysis is done by obtaining a "count" of "heard" or "quiet" samples. Based on the duration of the sample, it can be determined if the sample corresponds to a known cadence or if it would most likely be voice. By storing a history of both counts and the corresponding cadence, it is possible to accurately detect any of the possible tones (busy, reorder, dialtone...) and the occurrence of voice.
|
