Hi,
This circuit is really bogus for a number of reasons. It would be best to drop it entirely and start over.
Besides the other problems, it's not a good idea to use a transistor collector to cut an audio signal anyway, so the whole thing is not great to start with. The audio, if it is referenced to 0v (ground), only gets cut on the positive peaks which would produce really bad distortion. If it is referenced to some negative voltage then the non linearity of the transistors would still produce unacceptable distortion.
The modern approach would be to use DSP, but dated slightly before that the acceptable approach would be to use an LDR excited by some light source like an LED. As the LED is driven by the rectified and filtered audio, the LDR resistance decreases and thus cuts the audio. Since the LDR is a resistance, it shows the same conductance for both positive and negative signals and so cuts both polarities of the audio equally. The non linearity of the LDR is not with respect to the polarity of the signals, but with respect to the light level vs resistance which will not add any distortion as long as the time constant of the DC filter is longer than say 10 times the slowest audio signal.