The main criticism of the circuit by the magazine Elektor is that there is no device to know that a battery is completely out of order, short-circuit or reverse polarity, and it is possible that the load stops from a single cell is out of order, it is a fact.
So here's a more sophisticated system:
It uses a switching power supply, here a single integrated circuit, which boosts the voltage of a used battery at a high enough voltage to allow charging of one or more Ni-Cd or Ni-MH under a slow stream. The advantage of this system is that when the battery is fully used the switching power supply simply stops working ... so no problem when the battery is HS.
The built-in switching power supply circuit is a TL496 from Texas Instruments (datasheet:
http://www.datasheetarchive.com/preview/3915411.html )
It provides a DC volts 7,2 from 1,2 volts voltage of a battery. This voltage, rated for a battery, corresponds to that of a battery already quite used. According to the datasheet of TL496 the minimum input voltage is 1,1 volts, but in practice the TL496 continues to operate with a voltage 0,85 volts. Even with as low a voltage the circuit can still provide a current of several mA.
The output of TL496C provides the charging current with the resistor R which serves to limit the current and feeds a charging indicator LED with a low consumption 2 mA. The LED remains lit as long as there load. Both output circuits are separated by two 1N4001 diodes to the LED is turned on by the "used" battery, not the battery.