Author Topic: ( HELP ) a Complicated design for 16F877a (Read 1362 times)

credem · « **on:** May 19, 2009, 06:05:04 PM »

Guys, i am about interfacing 16F877a with a following feature listed below:

FACE RECOGNITION
VOICE RECORDER/PLAYBACK
EMBED to Payroll system (DTR)

does any1 has a circuit for face recognition and voice recorder? pls.. guide me on my first design.. tnx and regards to every1..

P.S: is this possible to create device for embedded system like this? 'coz, i dont have yet any idea for recognition and other mechanics for this.. i just dont want to include a PC interface with this face recognition except on sending its data as input for PC database... tnx again

Philippine Electronics Forum · « **on:** May 19, 2009, 06:05:04 PM »

marcelino · « **Reply #1 on:** May 19, 2009, 08:27:02 PM »

I'm not sure, but with facial recognition, a DSP chip (i guess) is better.
There will be a lots of mathematics - I again guess you can obtain those algorithms on the published journals (IEEE or ELSEVIER).

Philippine Electronics Forum · « **Reply #1 on:** May 19, 2009, 08:27:02 PM »

zer0w1ng · « **Reply #2 on:** May 19, 2009, 08:31:43 PM »

The PIC could not do face recognition.
And this needs a camera to capture faces thus needing interface for it.

Voice recording could be done if coupled with another storage chip.
Playback is possible with wav samples stored in flash.

For this app, I would suggest an embedded PC with a webcam attached.
This hardware is capable to do this. But software is another problem.
It is difficult to do. Another solution is to buy pre-made face-recognition libraries but is costly.

Philippine Electronics Forum · « **Reply #2 on:** May 19, 2009, 08:31:43 PM »

marcelino · « **Reply #3 on:** May 19, 2009, 08:34:59 PM »

ah... binibili po pala yun sir... hehehe

Philippine Electronics Forum · « **Reply #3 on:** May 19, 2009, 08:34:59 PM »

the_next_big_name · « **Reply #4 on:** May 19, 2009, 08:35:35 PM »

uu nga yung image processing module ng matlab is around 10k usd?

Philippine Electronics Forum · « **Reply #4 on:** May 19, 2009, 08:35:35 PM »

blackfin · « **Reply #5 on:** May 19, 2009, 09:32:55 PM »

16 bit processing is very slow for math extensive face recognition. you'll need at least 32 bit for the burden in processing load and pipeline manageable. you will need dsp processor for this. look for one with mcu and signal processing functionality.

rdpzycho · « **Reply #6 on:** May 19, 2009, 09:45:11 PM »

Quote from: credem on May 19, 2009, 06:05:04 PM

Guys, i am about interfacing 16F877a with a following feature listed below:

FACE RECOGNITION
VOICE RECORDER/PLAYBACK
EMBED to Payroll system (DTR)

does any1 has a circuit for face recognition and voice recorder? pls.. guide me on my first design.. tnx and regards to every1..

P.S: is this possible to create device for embedded system like this? 'coz, i dont have yet any idea for recognition and other mechanics for this.. i just dont want to include a PC interface with this face recognition except on sending its data as input for PC database... tnx again

sa complication ng mga dapat gawin, sa face recognition pa lang hindi na kaya ng 877, and in case malaki man ang memory niya, it may take minutes to recognize the features of a face. pang-DSP 'yung face recognition.

'yung latter 2, kakayanin with external storage. malamang Time In Time Out system 'to kaya may embed to payroll.

RiDdLeR??? · « **Reply #7 on:** May 19, 2009, 09:54:33 PM »

Mga pang 32 bit arm9 or blackfin processors na yata dapat kailangan diyan with embedded linux. Pwede din kaya yung mga 32 bit - 8 core propeller chips since it natively supports rgb video??

zer0w1ng · « **Reply #8 on:** May 20, 2009, 08:55:06 AM »

Quote from: marcelino on May 19, 2009, 08:34:59 PM

ah... binibili po pala yun sir... hehehe

Yes, to save time and complexities of a project especially for specialized algorithms such as this.
The theories are open and can be read at IEEE site but takes time and resources to make.
This will cost more for the developer than to buy the library.
And no need to re-invent the wheel.

leoren_tm · « **Reply #9 on:** May 20, 2009, 10:46:34 AM »

P.S: is this possible to create device for embedded system like this? 'coz, i dont have yet any idea for recognition and other mechanics for this.. i just dont want to include a PC interface with this face recognition except on sending its data as input for PC database... tnx again

IMPOSIBLE.....

anthonydp · « **Reply #10 on:** May 20, 2009, 12:37:44 PM »

To get you started with this microcontroller-based face-recognition project of yours, here is a great site that has much information about face recognition algorithms:

http://www.face-rec.org/algorithms/

I think this is do-able on an 8-bit mcu but it will be very **VERY** slow.
So, instead of doing this on a custom circuit, why not use an embeddable PC using these Atom-based motherboards:

http://www.rsun.com.ph/view_item.php?item=MOB015
http://www.rsun.com.ph/view_item.php?item=MOB017

All the circuits you need (for sound recording, data storage, 32-bit compute power, ethernet, usb expansion) are available in one circuit board. Plug in a DDR2 ram, a harddisk, a power supply, a usb webcam, and install Linux, and you are ready to go... program away...

Oh here's a link for PC-based face recognition goodies:

http://pages.cpsc.ucalgary.ca/~hanlen/vision/facelinks.html

By using a PC-based architecture in your face-recognition project, you'll have a higher chances of taking it to completion.

leoren_tm · « **Reply #11 on:** May 20, 2009, 05:14:28 PM »

it will be impossible on 8bit MCU(practically)...

e.novacek · « **Reply #12 on:** May 20, 2009, 05:53:31 PM »

i think much better to use MCU with DSP capabilities for Face Recognition..

Strong candidates are SHARC and Blackfin for Image Processing they have DSP engine co-processor inside.. TI and FreeScale offers a good line of DSP Processors.. DSC or dsPICxx ..ok din but it is more on Motor Control..

32bitters are good but the style of implementation is quite difficult to work with..

Usually this project is done via Matlab because the Matlab has the adequate toolbox to perform this job..
as well as match filtering.. u have to work with correlation, normalization and convolution for the algorithms..

and pretty much better if u have knowledge with wavelets, I-FFT and FFT.. it will help for sure..

marcelino · « **Reply #13 on:** May 20, 2009, 06:05:07 PM »

Quote from: e.novacek on May 20, 2009, 05:53:31 PM

i think much better to use MCU with DSP capabilities for Face Recognition..

Strong candidates are SHARC and Blackfin for Image Processing they have DSP engine co-processor inside.. TI and FreeScale offers a good line of DSP Processors.. DSC or dsPICxx ..ok din but it is more on Motor Control..

32bitters are good but the style of implementation is quite difficult to work with..

Usually this project is done via Matlab because the Matlab has the adequate toolbox to perform this job..
as well as match filtering.. u have to work with correlation, normalization and convolution for the algorithms..

and pretty much better if u have knowledge with wavelets, I-FFT and FFT.. it will help for sure..

hence... a complicated project.

RiDdLeR??? · « **Reply #14 on:** May 23, 2009, 06:22:06 AM »

Medyo OT, but speaking of DSP, what is the main component na meron ang DSP sa ordinary MCUs? I mean like dspics and pic24s, parehas din naman 16 bitter ang mga to.

blackfin · « **Reply #15 on:** May 23, 2009, 10:40:35 AM »

DSPs are much like CPU specially made for signal processing, has lots of peripherals, memory, ALU with special MAC or multiply accumulate core intended for intensive multi bit multiplication and addition which is the primary process of signal processing.

Quote from: hunyango on May 23, 2009, 06:22:06 AM

Medyo OT, but speaking of DSP, what is the main component na meron ang DSP sa ordinary MCUs? I mean like dspics and pic24s, parehas din naman 16 bitter ang mga to.

e.novacek · « **Reply #16 on:** May 23, 2009, 02:49:42 PM »

Quote from: marcelino on May 20, 2009, 06:05:07 PM

hence... a complicated project.

hehehe'

i don't see any good reasons why credem wants to create a complicated project.. if i'm to choose, i think

face recognition is enough.. but anyway..

credem.. if ur about to push-through this project.. my suggestion is planuhin u muna ng mabuti.. and solve

the easier problem first.. like for example ung payroll system.. unahin u muna un, then isunod u ung voice

recorder/playback.. step-by-step.. then from there u can make progress.. dahil kung mag-jujump ka agad

sa mahirap na task at first stage.. then ndi magiging positive ang result..

Philippine Electronics Forum · « **Reply #16 on:** May 23, 2009, 02:49:42 PM »

Author Topic: ( HELP ) a Complicated design for 16F877a (Read 1362 times)

Philippine Electronics Forum

Philippine Electronics Forum

Philippine Electronics Forum

Philippine Electronics Forum

Philippine Electronics Forum

Philippine Electronics Forum