Basic Functions and Model Selection¶

What is the process of developing products with Chipintelli’s ASR chip?¶

First, you can select the appropriate chip or module according to the relevant instructions in the document. If you are a beginner, you are recommended to directly purchase our development board kit for development. Then you need to register an account on the chipintelli speech AI platform, and make firmware or software development according to its use method to actually test the speech effect and product functions. If your products need to be mass produced, you can directly purchase the modules provided by our company in bulk, or design the circuit boards by yourself according to the hardware design data in the platform design data. Only after debugging, you can conduct the product performance test, reliability test and production test according to the test methods provided by the Document Center, so as to meet the requirements of the complete machine for mass production. If you encounter any problems during the development process, you are welcome to use the platform’s online support or send information to the company’s support mailbox support@chipintelli.com , or contact our technical support engineers, and we will provide you with comprehensive technical support services.

How to select the appropriate speech chip or module when developing products?¶

You can click ☞Hardware Selection Guide to obtain selection information. If there is your corresponding product in the product solution development document in the document center, you can directly use the suggestions in the document to select the appropriate chip or module for development.

The board shape provided by Chipintelli does not meet the requirements, can we provide customized design?¶

If you have the ability to design a board, you can purchase a ASR chip from Chipintelli and design it according to the hardware development document in the document center. If you need our customized board, we can also help you customize it when it reaches a certain order of magnitude. Please contact our business personnel for details.

Are there any recommendations for speech chips with 200 entries?¶

Our CI1103 ASR chip can support about 300 words. For the number of terms supported by different chips, please refer to☞Hardware Selection Guide.

How to select the number of microphones for speech products?¶

Our company has single microphone and dual microphone options, both of which support high recognition rate. The cost of single microphone is low, while the cost of dual microphone is high, but it supports sound source location, environmental noise suppression and other functions. Therefore, single microphone is recommended for general applications; Double microphones are recommended for loud noise and sound source positioning.

How to improve speech recognition?¶

There are many factors that affect the effect of speech recognition, such as the performance of the ASR chip itself, the choice of terms, and the structural design of the product (such as the installation position of the microphone and speaker). It is recommended that you directly use Chipintelli’s ASR chip products and speech AI platform for development. Our company has accumulated a lot of technologies and products, and the recognition rate of the standard scheme can reach 97%. You can also directly use the platform’s optimization and other functions to further improve the speech recognition effect.

How to support various dialect recognition?¶

Due to the variety of dialects, it is difficult to support them all at present. At present, our company has made special plans for some northern language family programs and some dialects that cover a large number of people. For some biased dialects, we can also use our local self-learning plan to support. Please check the details☞offline self-learning function .

What is the power consumption of Chipintelli chip? Does it support battery applications?¶

Because offline speech recognition needs to always enable the monitoring mode, its power consumption is currently difficult to achieve very small. For example, the standby power consumption of our speech chip CI110X (CI1102/CI1103) is about 10mA, and the normal power consumption is about 35 to 50mA. If it is used for battery application, considering the standby time, it is recommended to use a large capacity battery. Or if there is a key, you can use the key to turn on the speech function and then power on the speech module. When the speech function is not used, power off the speech module to save power. The chip scheme of our company can enter the recognition state after being powered on for about 0.5 seconds.

What are the advantages of offline speech over pure online speech?¶

The offline speech solution uses professional offline ASR chips to process speech at the terminal without connecting to the network, which can maximize the protection of users’ speech privacy, and can also be used after the network is disconnected, saving network costs and bandwidth. The offline speech scheme has good real-time performance and can basically respond within 1s. There is no delay problem in the cloud scheme due to network congestion or poor signals. It is very suitable for functional control products, such as large and small household appliances, cars, lighting and other fields.

What are the main advantages of the offline speech scheme compared with the pure online and pure offline speech scheme?¶

Off line solution is a combination of offline and online solutions. The basic operation control of the product adopts offline speech control, which has high security, online cloud services, and very rich content. Compared with the pure online solution, offline speech control is still available when there is no network or the network quality is poor, and the basic operation of the product is not affected by the network; Compared with the pure offline solution, the offline solution can provide consumers with rich cloud content services. This solution is particularly suitable for smart home appliances and smart home products that need to add content or services.

How to realize speech control linkage of smart home?¶

The speech AIoT scheme can be adopted, that is, the intelligent speech chip is used for speech function control, and WiFi, Zigbee and other wireless modules are added for interconnection between devices to achieve speech control of all connected smart home devices.

How to select WIFI in Chipintelli’s offline speech scheme?¶

The selection of WIFI for offline scheme shall consider the hardware interface used for transmitting audio, the main frequency of WIFI and the memory size reserved for offline scheme. The details are as follows:

WIFI hardware interface supports UART (above 921600 baud rate)/IIS SALVE/standard SDIO DEVICE.
The main frequency of the WIFI CPU is above 160Mhz, and the SRAM used for the offline application layer is greater than 200K Byte.
WIFI SDK supports basic protocols such as JSON, SPEEX, Mbedtls, MQTT, HTTP, etc.

For software design, please refer to the relevant contents of the software development SDK in the document center.

What cloud services can Chipintelli choose for its off line intelligent speech solution?¶

At present, our company has conducted docking tests. The cloud platforms that can be quickly supported include:

Tencent Cloud xiaowei AI speech.
Qi Wuyun’s AI speech.
Turing Cloud AI speech.
ROOBO cloud AI speech.

These cloud platforms can provide online AI speech function services for products, mainly including access to music, weather, smart home and other online content services through natural language interaction; Support speech recognition (ASR), semantic analysis (NLP), speech synthesis (TTS) core services, etc. Our company also has partners who support more cloud platforms. Please contact our business personnel for details.

Does the offline speech recognition scheme support the remote control function of the product, i.e. IOT function?¶

Our off line speech recognition scheme can add integrated IoT functions and connect IOT clouds (such as TUYA cloud, Tencent cloud, etc.) to achieve mobile phone remote control products. At present, our company also has mature speech AIoT product solutions for users to choose from.

Can I wake up the device normally when using sounds other than wake-up words?¶

The speech product uses wake-up words because if it keeps working normally, too many command words may increase the rate of false recognition. Therefore, one or a few words are required to wake up. Under normal circumstances, using sounds other than wake-up words cannot wake up the device normally. However, in order to give consideration to local accents in actual use, the wake-up word setting has a certain degree of ambiguity on the basis of ensuring the correct wake-up recognition rate. Therefore, it is possible that users can also wake up devices by sending similar sounds of wake-up words. In order to maintain better accuracy and compatibility, the design of wake-up words may need to be further optimized.

I have purchased a set of audio and development boards from your company. How can I connect them?¶

Please link the IIS signal cable on the development board backplane with the signal cable on the recording board, and use a mini-A cable or 5V on the USB to serial tool to power the development board motherboard, and use another mini-A cable to power the recording board, and connect it to the USB interface of the PC.

How do your modules generate speech firmware?¶

You can view the relevant usage content in the software development of this document center, and use the speech AI platform provided by our company to generate firmware.

How can your speech module customize the broadcast?¶

The broadcast can be customized through the speech AI platform provided by our company, which also provides speech synthesis.

Your speech module microphone is wired. Can you attach the microphone directly to the board? Which is better?¶

The welding time and temperature of the microphone solder joint will cause the sensitivity of the microphone to change. Therefore, it is recommended to use the wiring method. If the patch is installed, it is necessary to determine with the microphone manufacturer whether the microphone can be reflowed.

How to configure single and double microphone in the SDK of B03GS01S module?¶

This module only supports a single microphone, and the B03GT01S module is required to use the dual microphone scheme.

How are the microphones and speakers of the B02GS01S module connected to the module? Are there positive and negative requirements?¶

The “+” of the microphone corresponds to the left channel of the module. If the mic is installed reversely, the sound will be seriously reduced and the recognition rate will be affected. Please connect it correctly.

When the module supports a horn, there is no difference between positive and negative, and positive connection and reverse connection have no effect; If there are two speakers, one is left and the other is right, the connection cannot be reversed. The SPKL+corresponds to the positive of the speaker, otherwise the sound phase will be canceled, and there will be problems with the broadcast.

How do I configure the serial port on the SDK of the B03GS01S speech module?¶

Currently, the commonly used communication serial ports are uart1/uart2, user_ There are two macros in config. h, UART2_ ENABLE and UART2_ USEUART1_ ENABLE, when you need to use uart2, set the macro UART2_ ENABLE is set to 1; When you need to use uart1, set the macro UART2_ ENABLE and UART2_ USEUART1_ ENABLE is set to 1.

What principles should the speech recognition module follow when setting command entries?¶

The recommended command entry length is 4-6 words; Try to avoid using similar vowel sounds, such as kitchen/study, air drying/drying; Choose words with good pronunciation, such as “Pu”, which is difficult to pronounce aloud, and try to avoid it; Reduce words that are too colloquial or frequently used in daily life, such as “I”.

Does the firmware burning progress remain 0% or an error is reported?¶

There are several common situations when encountering such problems:

After pulling 3.3V on the PGEN pin of CI110X and CI112X chips, connect the serial port tool to the computer. After checking the corresponding serial port number in the upgrade tool, power on the device again to download correctly.
Measure the level of TX pin of the serial port tool. At present, the serial port burning level of CI110X and CI112X chips does not support data signals of 5V and above. If the level does not match, you need to modify the output level of the serial port tool or add a level matching circuit.
Check whether the power supply of the equipment is stable. Some use the serial port tool 5V to supply power to the equipment, which may lead to insufficient output power at the moment of power on, causing the equipment unable to communicate normally. It is recommended to use a separate power supply system to provide stable working voltage for the equipment.
Check whether the connecting wire is reliable. DuPont wire is commonly used to connect the serial port tool to the equipment end. Due to the long-term plugging and unplugging of DuPont wire, the metal connector is loose and oxidized, which reduces the stability of data transmission. It is recommended to replace the connecting wire and reuse it.

How to select the correct USB to serial tool?¶

At present, the serial port used for the burning and downloading of the offline speech chip of Chipintelli needs to support a maximum 2M stable transmission rate. It is recommended to recommend USB serial port conversion tools made by such highly stable serial port communication transmission chips as CH340G, FT232, etc. to reduce problems in the burning process and improve the burning efficiency.

How to solve the error warning during firmware packaging?¶

If an error warning occurs during firmware packaging, check the error message first. For example, after adding a code, the code volume increases Plus, if the original partition is exceeded, an error message will be displayed, “Partition User: The partition has exceeded the reserved partition.”. At this time, modify the “Reserved” in the User option according to the actual “Reserved” in the error prompt box, and refer to “New” every time when modifying 0x1000, fill in the correct parameters, and then refresh the address package firmware; If similar errors are prompted for other partitions, refer to the above method to modify the corresponding partition. (For more instructions, press the F1 function key in the tool to view the help file.).