Audio Signal Processing using TMS320C6713

[Divya Agrawal]

I have recently started using TMS320C6713 and am trying to start an speech recognition project for which i have to input 3 different audio words. I am not being able to understand in which format or how will i able to store this audio signal in tms kit so as to further process it. Any help would be highly appreciated.

 

[u7karsh]

To any controller, the language known is binary or hexadecimal. Thus we need a number system by which we can convert our floating point numbers to hexadecimal format.
Since your DSP is a 32 bit fixed point processor, we follow the 8.24 format of representing floating numbers.
The first 8 bits are for integral part, while the last 24 bits are for decimal part.
Of the first 8 bits, the first bit is reserved for sign, i.e., if it is 1, number is negative and if 0, number is positive.

Thus we can represent any number between -128 to 127 using this format.
To represent any floating number, the mathematics involved is:

8.24 Hex = Hex of( number_in_float * (2^24) )
Eg.
Number in float = 0.25
Hex of ( 0.25 * 2^24) = Hex of ( 4194304) = 400000
Similarly for -0.25 we get FFC00000

For the reverse process, i.e., to get float from 8.24 hex, use the below mathematics:
Float = (decimal of(number in hex))/ (2^24)

Eg 0x00A01456 gives 0.625310301780701


32 Bit Multiplication

If we multiply two 8.24 hexadecimal numbers, we get a 64 bit output. To convert this output to the desired 32 bit 8.24 format, remove the first 8 bits and last 24 bits of the obtained output


Hope this helps..
Utkarsh Mathur