First off, please note that the following schematic is only one way of accomplishing the construction of a peak detector. There are several modifications to my design that would greatly increase the practicality of the design (and from a business standpoint, would decrease costs and reduce manufacturing complexity). I’ll make note of these after I demonstrate the schematic and run through a brief explanation.

Below is one branch of my peak detector featuring a lowpass filter.
Lowpass Filter Branch

Working from left to right, we first come across a voltage source (PSpice was used to design and simulate this schematic). The voltage source represents signal coming from the positive wire from a headphone “splitter”. Typical input voltages from an iPod (volume slider at 50%) is in the range of 50mV.

The 180Ω resistor and 10μF capacitor comprise the passive lowpass filter.

Next in line is the pair of cascaded op amps. Both are acting as inverting amplifiers. Note that the first one has a gain of (-1)(160kΩ / 2kΩ =) -80. The negative simply indicates that the AC signal is inverted. The next op amp acts as a unity gain inverter (-1)(10kΩ / 10kΩ =) -1. This simply re-inverts the AC signal.
For this project I used 741’s biased with ±12V.

Finally, we have our LEDs (modeled as a stand-in 100Ω resistor since PSpice does not have a readily-available element for an LED).

Why is the gain so large?
The operational amplifiers need to be able to produce an output that can overcome the LEDs breakdown voltage. Typically, LEDs have a breakdown voltage between 2 and 3 Volts. So we’ll need to amplify the input signal significantly to produce noticeable flashes in the LEDs.

Why two op amps?
Two operational amplifiers were used because at the time of this project’s conception, I was just experimenting with operational amplifiers (and cascading these elements). The use of two op amps is actually very inefficient in this case. Another op amp is just another element to bias and another piece that could fail (unlikely, but saturation — much more likely — will also botch the peak detector’s performance). Also, since audio signals are AC, generally the positive half of the waveform is accompanied by a near-equal negative half. Thus, there is no need to re-invert the inverted signal. The LEDs will still flash, but their flashes will be a half wavelength out of phase. (This phase difference is undetectable by human perception.)
Assuming only one operational amplifier is used, the inverting op amp can be replaced with a non-inverting op amp.

To what frequencies do each channel respond to?
The lowpass responds to frequencies between 0Hz and 600Hz.
The bandpass responds to frequencies primarily between 600Hz and 1,800Hz.
The highpass responds to frequencies above 1,900Hz.

I have not provided the math for this, but if you’re dying to know, I can provide this.

How can I adjust the frequency sensitivities of each branch?
To do this, you’ll need to adjust the respective inductance, capacitance, and resistance values of the filters. Again, I’ll provide or point to the math if desired.

Here are the branch schematics for the lowpass, bandpass, and highpass configurations. These schematics show my first design (with the unnecessary op amps).
Lowpass Schematic
Bandpass Schematic
Highpass Schematic