DV (Digital Voice) using PSK at 93 bps The new PSKDV Digital Voice soundcard mode allows acceptable voice quality over quite bad HF channels. It uses a new type of voice compressor with extremly high compression rates, and a transmission system very similar to psk31, but 3 times faster.
The voice compressor does not try to extract sound parameters of a vocal cord / tuned pipe model like lpc or other low bitrate codecs. Instead it tries to extract harmonic magnitudes for vocals, and spectrum transitions for consonants. Then it tries to correlate those parameters with a table of 78 phonems. The extracted phonems and quantizized pitch and duration parameters are then sent over HF at 93bps. First tests have been very sucessful at SNR's down to -4 db. The mayor problem seems to be that the current phonem table is tuned to german language, and gives poor results in those strange foreign languages. Another issue is that you wont recocnize the voice of your qso partner. Individual voice aspects are lost in the compression process. Only pitch and talk-speed are preserved. The beta test program is downloadable in the usual place, but only today, 1. april.
