Classification of voiceband data signals

The viability of a voiceband signal classification technique which can service as a front-end for digital signal interpolation is demonstrated. The technique uses second-order statistical parameters of the complex envelope of the input signal. The input signal is assumed to be in μ-law PCM (pulse code modulation) form. For observation windows of at least 128 ms, it is shown that this method correctly classifies the input baud (e.g. 2400, 1600, 1200, 600 and 300 symbols/s) and modulation type (e.g. QAM, PSK, and FSK). This information allows the determination of the bit rate of most voiceband data signals. Coupled with a digital signal interpolator and variable rate coding, an increased compression ratio on long-haul digital trunks can be achieved