I’ve done this very thing using a tracker and Trapcode’s soundkeys. You can set the output of the waveform to include only the Y axis. Then for further refinement you can divide the expression by any number to soften the expression’s effects.
In your position expression, create a variable of “temp=this.complayer”. When you create your array enter 1 for x (ex. [1, temp, temp]. Now your variable will only be effecting the y and/or z values.