Local Voice USB Controller

The main control chip uses Qiyingtailun's single-microphone offline speech recognition module CI-B02GS04T (1102 golden finger board), Lichuang Mall No. C2693426. This module only needs to use a single microphone to achieve 10-meter ultra-long distance speech recognition, and the recognition rate can reach more than 97%. And it has 3 UARTs, 6 PWMs, and 1 I2C interface, with a total of 26 IOs available. It is a low-cost, high-performance, high-reliability offline speech recognition module.
 
The LED driver chip uses the RY3730 chip of RYCHiP (Ruiyuan) Company, Lichuang Mall No. C370887. RY3730 is a boost converter suitable for driving 8 white LEDs from a single lithium battery. The use of 250mV feedback voltage reduces the power loss of the entire circuit and improves efficiency. The optimized operating frequency can meet the requirements of low standby operating current and high efficiency. This chip has the same working conditions as the AP3032KTR of DIODES (US and Taiwan) Company, and is half the price.
 
The acrylic panel is drawn with the Easy EDA panel editor and produced by the Easy panel platform. The effect and accuracy are excellent, economical and affordable.
 
Preview of the finished product:
The hardware circuit part
 
is connected to TVS at the microphone and speaker to prevent static electricity from affecting the device, as shown in the picture:
It should be noted that the offline voice module uses 5V for power supply, but the communication and GPIO parts are It only accepts 3.3V level, so in the burning program or GPIO control part, if the external signal is 5V, level conversion needs to be used to prevent damage to the module.
 
For the LED driver part, you can directly use the official basic application circuit, as shown in the figure below. R1 is the feedback and leakage resistor. It needs to use a high-power package such as R2010 to avoid causing high heat. The resistance value is 0.25/the total working current of the LED. inferred.
 The PG_EN pin of the module is a programming pin and needs to be connected externally to facilitate the programming process.
The offline voice module uses the gold finger version, and the matching gold finger female socket cannot be found in the mall, so the module is installed on the base board by welding. Here you need to pay attention to the order and spacing of the gold fingers to prevent installation. Problems with misalignment or failure to weld.
 
In the software part,
 
the first step is to obtain the SDK. Register and log in to Qiyingtailun Voice AI Platform, click on the data download on the left, and select the corresponding data to download. This module uses CL1102, so choose to download this version:
the second step of model production. Select the language model and broadcast sound synthesis on the left respectively, fill in the corresponding form according to the prompts and upload it to the platform for recognition, download the recognized model file and replace the default SDK content.
 
The third step is to download the IAR software, open the file ending with eww in the project folder in the SDK, and you can see all the project files. First click project->clean in the upper left corner, then click make, and wait for the platform to compile all the codes. If it displays 0error, that's it.
 
Open the user_msg_deal file, locate the uint32_t deal_asr_msg_by_semantic_id function, and you can see the part that responds to the asr message processing according to the semantic ID:
the number following the case here is the ID number following each command in the [60000]{cmd_info}.xls table , adjusting the actions performed by each command according to the table can realize operations such as turning on and off lights by voice.
After editing all the content, click make to generate the BIN file.
 
The fourth step is to open the synthetic partition bin file .bat in the firmware folder, select 1 and wait for the window to close automatically after synthesis, then open the package upgrade .bat file,
select firmware packaging -> package firmware, it will prompt that the firmware has been generated, and then click Firmware Upgrade button, connect the motherboard and programmer, and short PG_EN and 3.3V. Disconnect the power supply of the connected module again, the progress bar will start to load, and after waiting for about 3 minutes, you will hear the sound of the device starting up. You can test whether the night light and USB interface are working properly through various voice commands.
 
To achieve the effect,
 
voice turns on and off the night light, "dark" and "bright" adjusts the brightness, "turns on the switch" and "turns off the switch" adjusts the USB output. The USB uses thick wiring and can be connected to USB fans, desktop charging stations and other devices. .
 
ps: This device is going to be given to a roommate who studies medicine in the university. When he was a freshman, he dreamed of becoming a dean, so the reply here is Dean Hu to realize his dream in advance.
 
The demonstration video has been uploaded to the attachment:
 
 
 
 
PS: The firmware and SDK have been uploaded to the project attachment. You can download and change it, or you can burn it directly. If you have any questions, you can contact me via private message for help.