Most ASV systems use 8 bits and around 8KHz sampling frequency for their input.
The largest reason for this is probably that most research is performed on telephone lines, where
the bandwidth is even more limited, so sampling at more than 8KHz would give minor improvements.
However, some systems would perform better using a higher sampling rate. This indicates that better
performance still can be achieved with higher quality of data.
Data stored in A-law(电话信道的语音编码方式) format use a non-linear scale(压缩), and 8 bits can be used to represent a 12-bit linear dynamic range, with acceptable impact on detail resolution.
To accomodate for the higher dynamic range, higher values are more widely spaced than are the lower
values. This results in the preserving of both fine details and a large dynamic range. It is a common
representation for speech.