Part Number: TMDX5535EZDSP
Tool/software: Code Composer Studio
I am trying to follow the step by step instruction on this document to build and test the speech recognition software but I get many errors.
Here is the link to the document:
When I try to build the the TIESRdemoC55 then I get the following errors:
<Linking>"../C5535.cmd", line 18: error #10263: MMR memory range has already been specified"../C5535.cmd", line 18: error #10264: MMR memory range overlaps existing memory range MMR"../C5535.cmd", line 19: error #10263: DARAM0 memory range has already been specified"../C5535.cmd", line 19: error #10264: DARAM0 memory range overlaps existing
Do you have any solution? What is the correct step?
In reply to Lalindra Jayatilleke:
Yes, all I am doing is capturing audio on the Bluegiga and passing it to the C5535 via I2S1 for the recognition.
Since Bluegiga clock is at 16 KHz then I thought to follow the steps in this Forums. I just wanted to make it to work with Codec I2S2 first before connecting the Bluegiga.
When I followed the steps except (changing the codec gain), the voice recognition doesn't work anymore.
Would you please let me know how to make the changes first that works with Codec at 16KHz?
Thank you for your support.
In reply to farzin dinyarian64:
No, I didn't make any progress.
Our current setup with BlueGiga is at 16Khz, I can't do at 8 KHz.
Would you please just let me know if these changes work with current setup? I made the changes but it doesn't recognize anything.
As mentioned earlier, I'm not sure how well TIesr would work at 16KHz since its not supported in the default demo.
However, let me try to assist based on the Sensory (third party) demo that we typically recommend customers to build their voice trigger systems with. Attached is a .out file that you can run on the ezdsp5535 assuming you have the I2S1 lines connected to your secondary device that does the audio capture. This .out was tested on my end with a C5517 EVM that has a linear microphone array connected. The audio captured on the C5517 is then be piped (via I2S) to the ezdsp5535 for the keyword recognition. In this case, the keyword is "Hello Blue Genie".
The Sensory library is provided as a 30-day trial at http://software-dl.ti.com/dsps/dsps_public_sw/dsps_swops_houston/C55X/Sensory-THF/latest/index_FDS.html Contact Sensory if you need to use this library in your product and they can work with you on the licensing. http://www.sensory.com/
Below is a diagram of the demo. A video is also attached. This will at least enable you to show-case the keyword recognition on the ezdsp5535 with audio coming from your BlueGiga device. Keep in mind that the audio will have to be presented to the ezdsp5535 as 16KHz sampling. I2S1 on the ezdsp5535 is SLAVE.
In the video, you will see that the P3 header was used on the ezdsp5535 to connected the I2S1 lines. You can also use the edge connector P2. Note the way the ezdsp5535 is armed by pressing the SW2 button once the .out is loaded. Please excuse the untidy wiring of the HW pic, did not get a chance to clean it up.
Please try this .out in your system.
Thank you so much for your last update. I actually tested the Sensory code with my Setup BlueGiga connected to TI over I2S1 and it worked great.
I am just wondering if you could compare the sensory code with TI code and let me know what are the changes to make it working at 16KHz. I am not sure if we could use Sensory solution at this time but we are very interested in using TI solution if it supports 16KHz. The TI documentation suggests that it should support 16KHz but it doesn't work when I change the sampling rate.
I would really appreciate your support. It would be a great Holiday gift for me.
Thank you for sending the binary image of the Sensory implementation on the CC5535 EzDSP on the Speech Recognition EVK. I have a few questions regarding the I2S waveforms.
The waveform below is from the unmodified 8KHz implementation showing the I2S waveform between the Codec and CC5535 chip.
As you can see, both left and right channels contain audio data. If the audio stream is mono, why is there data in the Right channel? Also, each channel contains 32 bit-clocks. Is the audio data 8-bit or 16-bit? Based on where I think the LSB is, to me it looks like 16-bit data? Is this correct? Where is the MSB? Should it be the left-most bit, one-bit delayed after the falling-edge of Word-Select?
The below waveform is from the modified 16-KHz implementation feeding the Sensor speech recognition engine in the CC5535.
In this case, there is once again, audio data in both the Left and Right channels. Also, there are 24 bit-clocks in each channel instead of 32 as in the previous example. Once again, is the audio data 8-bit or 16-bit in this case? If the MSB is one bit delayed from the falling-edge of Word-Select, then it appears there are 16-bits per audio sample. Is this correct?
Thanks for your help on this and Merry Xmas and Happy new year.
In reply to Mahsa Sasani:
Would you please attach your project that builds on your Linux Ubuntu system including the Make File?
Would you please also send me the PATH that needs to be included?
All content and materials on this site are provided "as is". TI and its respective suppliers and providers of content make no representations about the suitability of these materials for any purpose and disclaim all warranties and conditions with regard to these materials, including but not limited to all implied warranties and conditions of merchantability, fitness for a particular purpose, title and non-infringement of any third party intellectual property right. TI and its respective suppliers and providers of content make no representations about the suitability of these materials for any purpose and disclaim all warranties and conditions with respect to these materials. No license, either express or implied, by estoppel or otherwise, is granted by TI. Use of the information on this site may require a license from a third party, or a license from TI.
TI is a global semiconductor design and manufacturing company. Innovate with 100,000+ analog ICs andembedded processors, along with software, tools and the industry’s largest sales/support staff.