A Pure Offline Speech Scheme for Serial Port and Electronic Control Communication¶

New users are recommended to check the ☞novice guide to understand the basic knowledge of development, and then view this document

1 General¶

Offline voice is used more and more in various products, which adds “ears” and “mouths” that can hear and speak to the original non intelligent products and improves the user experience. The most common way to add offline speech recognition in applications is to use serial communication to communicate the serial port of the voice chip with the serial port of the MCU of the user’s electronic control board. The advantages of this method are as follows:

Rapid development and verification; *Flexible application: This method can be flexibly applied to multiple products, and different products only need to modify the serial port protocol and voice command words.

Our company has specially prepared standard modules for such applications, which can guarantee the rapid shipment of products after users directly select products.

2 Scheme Advantages¶

At present, our company has iterated multiple generations of chips, which has strong advantages in the offline voice field. Some features of our voice scheme are as follows:

Fast identification response: generally completed within 0.2~0.8s;
High recognition rate: recognition rate>95%;
High reliability: The chip scheme has passed the double 85 and EMC tests, which can meet the high quality requirements of Gree, Midea and other appliance giants;
Mature data: complete software, hardware and structure design scheme data can save development cycle and reduce development difficulty for users;
Convenient customization: We can use our voice AI platform for convenient development, and we can quickly and flexibly customize wake-up words and command words without requiring users to provide corpus;
Ultra far recognition: it can still be recognized when it is 10 meters quiet.

At present, our voice solutions have been widely used in mass production in various fields, such as fans, clothes poles, heating tables, tea bar machines, lighting, etc.

For more information about our chip solutions, please contact us through the following ways.

Tel.: +86-028-61375925 Email: support@chipintelli.com

3 Scheme introduction¶

If you are first exposed to voice solutions, it is recommended that you first look at the complete product design process ☞product solution development process.

This part focuses on the block diagram, modules and structure considerations in the scheme design.

3.1 Scheme block diagram¶

The design block diagram of the scheme is as follows (taking the clothes pole as an example). The voice module and the electronic control module need four serial ports to connect. The voice module plays the voice according to the state of the product’s electronic control, and the product’s own control is still completed by the electronic control module.

The voice module and the electronic control module exchange information through the serial port. There are two main ways. One is to play the voice module actively and generate the information to the control module at the same time; The other is voice module passive play, which is described as follows:

3.2 Serial communication mode¶

Voice module active play

The active play mode mainly performs the following actions:

Recognize entries;
Actively play the feedback sound;
Inform the electronic control module of this message serial port;
The electronic control module executes actions.

Advantages: simple control mode, one-way output information from voice module to electronic control module.

Disadvantages: It is impossible to give voice feedback according to the current situation of the product and the keys controlled by the electric control panel.

Voice module passive play

The passive play mode mainly performs the following actions:

Speech module recognizes entries;
Inform the electronic control module of the entry information serial port;
The electronic control module executes relevant actions;
Inform the voice module of the content to be played;
The voice module plays according to the feedback of the electronic control module.

Advantages: The voice module plays the corresponding content according to the status of the electronic control module, which can provide the most appropriate feedback and have a good experience; The electronic control module is controlled by pressing keys or remote control, and the voice module can also play the status.

Disadvantages: The electronic control MCU chip requires a small amount of code development work, and the development cycle is slightly longer than the active play mode.

3.3 Hardware selection¶

For this scheme, our company has a number of standard modules that can be supported. Users can click the ☞Hardware Selection Guide to view and select the appropriate modules. Here are some modules that can be used for serial communication, as follows:

3.3.1 CI130X module¶

Such modules include CI-D02GS02S, CI-D02GS01J, etc., as shown in the figure below:

matters needing attention:

If the voice module board is powered by the electric control panel, note that the power supply capacity of the electric control panel to the voice module needs to be greater than 500mA;
The level of CI-D02GS02S serial port supports 3.3V and 5V, which can be configured through software. Note that when 5V is used, the backplane needs to add a pull-up resistor, and the software needs to enable macro definition: UART_ PAD_ OPENDRAIN_ MODE_ EN；
When using CI-D02GS02S, be sure to reserve UART0 and PGEN contacts or sockets on the backplane for subsequent online upgrades;
CI-B02GS06J has pull-up resistance on the PCB, and only supports 5V level by default.

3.4 SDK package description and module serial port communication protocol¶

Our company has provided users with a complete SDK package for this solution. Users can go to ☞Chipintelli Speech AI Development Platform.

It is recommended to use our standard protocol for serial communication of the module. For the specific format, please refer to ☞Serial Protocol.

Note:

It is recommended to use different acoustic models in different fields. The details can be seen in the list when making models, as follows: (Note the code configuration suggestions for related models);

The default parameters of serial communication are: * Baud rate: 9600 * Stop bit: 1 * Data bits: 8 * Parity: None
What commands and functions are obtained by judging “Command+Voice ID” in the code;
It is recommended to check the integrity of the sent serial port information.

4 Product Structure¶

In the offline voice solution, the product structure design is very important, and the location of the microphone and speaker will directly affect the user experience. This part focuses on the structural design of the microphone and speaker to ensure the identification effect of the finished product.

4.1 Precautions for microphone structure design¶

It is recommended to choose an omnidirectional analog microphone with a signal to noise ratio of more than 70dB and a high cost performance of - 32dB ± 3dB;
If the vibration of the product is relatively large, select a thicker microphone rubber sleeve to reduce the vibration, which can reduce the impact on the recognition effect;
In some application scenarios, it is easy to accumulate water or dust, so it is necessary to select a waterproof and dust-proof microphone (provided through communication with the microphone manufacturer);
The microphone head is a sensitive device, and it is prohibited to directly spot hot sol;
Insert the microphone completely into the fixing hole to avoid deflection, and align the center hole of the pickup hole with the center of the microphone;
Select RTV silica gel that meets the environmental RoHS requirements. The recommended common silica gel is 703/704/737, or other organic materials, single component room temperature curing silica gel, etc;
Most silicone rubber grouts with a thickness of less than 3mm can be cured for 8-12 hours at real temperature, and those with a thickness of more than 3mm can be cured for a longer time, which can be poured in layers for many times to ensure complete curing, stability and sealing;
The production process can advance the process of fixing the microphone. In addition, pay special attention to the fact that the microphone wiring is pulled by gravity, which may cause the microphone to be pulled askew before it is fully solidified. Also, pay attention to fixing the microphone plug cord.

4.2 Precautions for horn structure design¶

Considering the sound quality, the speaker power should not be too small. It is recommended to use a speaker of more than 1W. The speaker with a sound cavity will have a better effect. At the same time, note that the power amplifier drive capacity on the hardware board should match the speaker model;
The speaker is mainly placed in a beautiful and convenient structure layout, but it should be avoided to be placed in the same position with the microphone as far as possible;
The speaker without sound chamber can be sealed with a sound chamber on the structure to increase the volume output.

The following figure shows a speaker specification and opening suggestions.

4.3 Other precautions¶

Taking the structural design of the clothes pole as an example, the following matters should be noted for the microphone:

The microphone layout shall be far away from the noise of the clothes pole and the motor;
The microphone shall be placed at the front/bottom of the clothes pole to avoid blocking and ensure the maximum range of pickup;
When the decibel meter is running, the steady noise at the microphone shall be less than 60dB;
If the microphone connection is long, the shielded line can be selected;
The microphone cable shall be bundled away from the motor or electric control cable.

The following figure takes the lifting clothes pole as an example to describe the suggested opening and position of the microphone structure. The microphone wire should also be considered for the convenience of leading out the wire and avoiding interference for the structural members of the fixed microphone.

5 List of relevant reference materials¶

No.	Reference description	Access method
1	Standard offline SDK CI110X_ SDK_ ASR_ Offline	Please go to ☞start Chipintelli Speech AI Development Platform
2	Standard offline SDK CI112X_ SDK_ V1.4.1	Please go to ☞start Chipintelli Speech AI Development Platform
3	Standard offline SDK CI130X_ SDK_ Offline_ 1.2.7	Please go to ☞start Chipintelli Speech AI Development Platform
4	Selection guide of our chips and modules	Please click the ☞Hardware Selection Guide to view
5	Reference of Huawei’s chip hardware design method	Please click ☞hardware design reference to view
6	Model selection of microphone and speaker devices	Please click ☞Compatibility list of peripheral devices to view
7	Product structure design	Please click on ☞Product structure design to view
8	Software development method and standard SDK description	Please click ☞Software development to view the contents of each part
9	Product identification performance test method	Please click ☞identification effect test to view
10	Method of production test	Please click ☞production test to view

If users encounter problems in use, please submit technical work orders onChipintelli Speech AI Development Platform, or directly contact our technicians for assistance.