voice enryption

[jassi]

hi,
i am in the final semester of my engg and my team(4 members) is tasked
to do a project on encrypting voice of normal radio commn.for this we
r reqd to make a module or attachment which would take audio as input
and do the processing and give output back to the rodio set .the
module is reqd to convert Analog to digital...do
compression ...encryption...back to analog for radio to transmit...
also FEC and Synchronisation is the major issue.we r to use XILINX
fpga...XC2S100 as this is available in the lab..

1.can u suggest an approach and feasibility to this?
2. what devlp board will be reqd for fpga and vhdl info?
3. can all this be done on fpga itself? i am clueless about what fpga
is capable of....

thanx
jassi
india

 

[Frank Raffaeli]

hi,
i am in the final semester of my engg and my team(4 members) is tasked
to do a project on encrypting voice of normal radio commn.for this we
r reqd to make a module or attachment which would take audio as input
and do the processing and give output back to the rodio set .the
module is reqd to convert Analog to digital...do
compression ...encryption...back to analog for radio to transmit...
also FEC and Synchronisation is the major issue.we r to use XILINX
fpga...XC2S100 as this is available in the lab..

1.can u suggest an approach and feasibility to this?
2. what devlp board will be reqd for fpga and vhdl info?
3. can all this be done on fpga itself? i am clueless about what fpga
is capable of....

thanx
jassi
india

My assumption is a "normal" radio comm has a base-band response of 300
Hz to 3.3 kHz.
Digital voice, for decent quality, will require some compression. You
can go one of two ways:
1) Frequency-domain compression
2) Time-domain compression

For frequency domain, research techniques like trigonometric
transforms (DCT), and also see FFT.
For Time-domain, see ADPCM.
I have another suggestion for time domain: Try encoding and decoding
the audio in NICAM format, such as what is used in Stereo TV
broadcasts in UK, HK, Sweeden, Spain, France, etc. (Nearly
Instantaneous Companding Audio Multiplex) It won't fit in a "normal"
radio channel; however, you can adapt the algorithm to slow down the
rate. It would still be enough for voice.
The NICAM algorithm will take care of the frame synchronization and
some basic error detection.
Here is a reference:
http://www.distel.co.uk/data_nicam.htm
If you can't find the ITU spec, I might have a copy.

Frank Raffaeli
http://www.aomwireless.com/

 

[Wim Ton]

hi,
i am in the final semester of my engg and my team(4 members) is tasked
to do a project on encrypting voice of normal radio commn.for this we
r reqd to make a module or attachment which would take audio as input
and do the processing and give output back to the rodio set .the
module is reqd to convert Analog to digital...do
compression ...encryption...back to analog for radio to transmit...
also FEC and Synchronisation is the major issue.we r to use XILINX
fpga...XC2S100 as this is available in the lab..

1.can u suggest an approach and feasibility to this?
2. what devlp board will be reqd for fpga and vhdl info?
3. can all this be done on fpga itself? i am clueless about what fpga
is capable of....

Have a look at the GSM algorithm.
What is your bandwidth? If it is SW radio (about 3 kHz), then a DSP sounds
more suitable than a FPGA for the required agressive voice-compression. I
suppose a FPGA is ideal for FEC and crypto.

A common approach is to used 2 layers of FEC:
You add FEC to a block of data.
Then you take a few bits from each block and put them in a new block with
FEC and send them over the radio-link.
The purpose: if a whole 'upper' block is lost, then it will affect only a
fews bit per 'lower' block, which can be corrected by the 'lower' FEC.
Single bit errors are corrected by the 'upper' FEC. The disadvantage is of
course a long delay between sending and receiving

For crypto, you use a stream algorithm. AES is said to be feasable in 10-30
kgates. If you are tight on gates, a self shrinking generator.

B.t.w. an encrypted link with 2.4 kb/s sounds like a robot with a serious
cold.

Wim