2012-05-04 06:45 AM
Hi All!!
Do you think
this
is possible
in the
Controller
speech recognition
or
speech
-driven applications
to prepare
?
2012-05-08 01:45 AM
That Microchip link is a good example for what I tried to suggest. It is a specialized solution, that obviously works well for the specific and narrow purpose it's designed for. As an example, it is fixed to english. If your project conditions match with the specs of this solution, it's okay. If not, you will have a problem with it.
That lib uses obviously scaled integer arithmetics. It is really fast, but very unflexible. With an M4F you have the capability to have a more flexible solution - if your project budget and roadmap allows for that, and you have any use for it at all.2012-05-08 10:54 PM
Yes.
I
think the same thing
.
Thus
the
principle
tool
for this.
2012-05-09 01:50 AM
Generally, the idea of realizing speech recognition with a Cortex Mx sounds really interesting.
When optimizing for an architecture and abandoning a generality approach in your code, an astonishing performance is possible. About 20 years ago, I have seen a Z80 based measurement system clocked with 2.5 MHz, doing a FFT based real-time audio analysis.2012-05-11 07:14 AM
I also
tried
to solve
Z80
device
.
(Sin
/ cos
ROM
table
for
the FFT
)
There was
100%
.
I
already
had
high-speed
CPU
(
4MHz
) :)
I hope
this
will be
F4
MCU
with more
success
.
2012-05-11 08:44 AM
> There was
100%
.I
already
had
high-speed
CPU
(
4MHz) :) Not sure what you mean. But that z80 was the 'original' design, 4 clock cycles per instruction. The sin/cos values were for sure tabulated. And, the FFT was obviously solely based on integer calculation. This was an example of tricky design & implementation to the hardware. A Cortex M4F is assumed give you the possibility to use FP numbers. With a FPU you don't gain anything with an integer FFT, but loose either precicision or you are burdened with higher complexity.2012-05-11 12:13 PM
Hi all,
I think the
integerFFT
onthe
cortex
M4
cangive
an incredible
performance
gains
.On
the
cortex
m3,
with the ST dsp library
, the result are alreadyremarkable.
The
cortex m4 core have
most powerful
DSP-like
instruction
implemented, so a integer FFT implementation can give surely better performance.
2015-02-07 05:13 PM
Playing about with a simple 'USB voice keyboard' using an STM32F4 Discovery board and a USB cable:
https://www.youtube.com/watch?v=rNjDnielkM8&feature=youtu.be
2015-02-18 01:07 PM
2016-10-24 12:07 PM
Hi,
I'v found that document on ST's webpage.http://www.st.com/content/ccc/resource/sales_and_marketing/presentation/product_presentation/30/8f/78/6a/b9/a4/45/ab/stm32-stm8_embedded_software_solutions.pdf/files/stm32-stm8_embedded_software_solutions.pdf/jcr:content/translations/en.stm32-stm8_embedded_software_solutions.pdfPlease look at page 48. It says Sensory ''speech recognition'' library is available. Does anyone know where to get it?2016-10-24 01:01 PM
Does anyone know where to get it?
With all such things, if it is not available for immediate download, contact your local ST Sales Office.