Voice recognition part was pretty tough to be implemented on-board. After wasting days to re-inventing the wheels by trying to implement the AVR based voice recognition protocol. But it had many limitations like Speaker dependencies, Limitations of wordlengths, Bad efficiency. Also I am not an expert in signal processing so i directly bought an EasyVR module which is interfacable by UART protocol. Library was available for Arduino thus had to port it to AVR's (Posted it on my blog check links)
Measurement unit is designed in such a way that the range need not be made manually, the uC will do it itself. (will post the circuitray soon)
Atmega128 is used as its brain, the main reason was its huge 128KB of flash. For synthesising speech i stored multiple blocks of 0-9 , point, V,F,Ohm,mA and it took almost 72KB memory for almost all these files. right now a 8ohm speaker is connected through an LPF but will soon switch it to a proper Audio Amplifier