Objective evaluation of the audibility of transient errors in an adaptive A/D conversion channel

Investigation of an AGC for audio applications

“…However some of the test subjects have been able to reach into the minimum gain step size implemented in both test (0.01dB). In [6] it was claimed that for making the transient errors (glitches) to be inaudible, one can target ADB and MFPD for a specific range, which is mapped to step sizes larger than 0.01 dB step size, the reason for this claim is that in [6] the number of steps has been maintained while the step size has been reduced, although we have followed a different logic, which is reducing the step size while having more steps, therefore our results are different with [6]. Consequently we cannot use ADB and MFPD as the objective measures to make any conclusion as the glitch is audible where the ADB and MFPD are not able to evaluate the audibility.…”

“…1.b). A Tuba music sample from the EBU Sound Quality Assessment Material CD [9], which lasts for 2.5s was chosen as the input of the channel (as it was reported to be the worst case scenario with regards to the level of the transient errors (glitches) based on [6]). For generating the sound files, initially each input signal was up-sampled by a factor of 8 from 44.1 kHz into 352.8 kHz to avoid the aliasing of the harmonic distortion which is the result of the AGC activity.…”

“…This method estimates the audio quality of the signal by incorporating the human auditory system properties. For our specific application (evaluation of the glitches audibility) the two recommended MOVs (Model Output Variables) of PEAQ [5,6] that are ADB (Average Distorted Block) which returns the logarithm of the ratio of the total distortion to the total number of severely distorted frames and MFPD (Maximum Filtered Probability of Detection) which measures the maximum of the probability of detection after low pass filtering, are utilized. These two objective metrics are used to Glitch assess the transient error level of the signals.…”