Intelligent Telephone robot, using Microsoft Speech Recognition Technology (speech SDK)

Recently, the company has taken over a challenging project. It is very interesting to share with you some of the gains and insights in the development process.

Intelligent Telephone Robot

The original committee is like this. There is a big guy with a lot of hands. In short, there are a lot of mobile phones with X cards. He wants to recharge these X cards to different user mobile phone numbers. Some friends may say, "You can talk to mobile, get mobile interfaces, develop programs based on interfaces, and then add the X-card password to the database, read with the program, and then call the mobile interface to complete recharge." Well said, pretty! In addition, when the X-card password is added to the database, the boss has already found someone to work overtime all day and night, all of which have been written into the database.

However, the fact is miserable. For various reasons, it cannot be done through the mobile interface. You can only use 1xx86 voice calls. As a result, we have the need for this project.

The requirement is not complex, that is, a telephone robot is required to simulate a human, call the phone, and press the corresponding button to enter the specified menu or enter some data (for example: mobile phone number, X-card password), and finally the most critical, you need to identify the other side of the prompt, whether the operation is successful or failed to determine whether the operation is successful, and record the operation results in the database.

 

The following goals must be achieved:

• You can use programming to control the device to complete the operations such as host picking and dialing until a service phone is called (for example, 1xx86)
• Ability to perform speech recognition to identify the other party's keywords on the phone (for example, Press 1 for Mandarin)
• Perform different dialing operations based on keywords (such as Mandarin), for example, by 1
• Enter the required data correctly, such as the mobile phone number and the X-card password.
• Identify whether the operation is successful Based on The Voice of the other party (for example, recharge successful)
• The data required for the operation and the records generated by the operation are stored in the database.
• The program must be robust to cope with various exceptions. For example, speech recognition is not possible, or it is not accurate, there are misjudgment, the telephone line quality is not good, and the phone cannot be accessed.

Implementation Scheme

Based on the above requirements, I found that two problems are important:

1. What devices are used to complete dialing, buttons, and other interactions with 1xx86?
2. What technology is used to implement the speech recognition 1xx86?

To address these two technical problems, I did a technical survey. Question 1: What equipment is used to interact with 1xx86?

I first thought of voice IVR and did some research, but found that this is not what we need. The main function of IVR is to act as the "server" of a voice phone, that is, the called party. When the user calls the phone, the IVR plays the voice to the user and then performs different operations based on the user's buttons.

That is to say, IVR acts as the called party and responds. This is obviously not what we need. What we need is to take the initiative to call 1xx86 as a "client", that is, the caller, and then press the key to select a different menu, or enter the mobile phone number and other information, you do not need to play the voice or recognize the buttons. What we need is the ability to dial and perform button operations.

To sum up, IVR is passed.

Then I came up with another solution, modem !!!

What? Did you think of an old-fashioned cat with a dial-up 56 K? Yes, but what I need is a programmable modem that supports the AT command. I searched the internet and found some devices that really sell them. They call them industrial modem.

Just do it, and find the sales result and purchase an industrial modem. Something is very inconspicuous, an iron box with many plug-ins on it: power AC, serial RS232, microphone mic, Horn SPK, telephone phone, telephone line.

I need serial RS232, speaker SPK, telephone line.

Use an audio cable to connect the modem's SPK port to the PC's sound card line to enter the line in interface. (Note that the line must be line in and cannot use the mic interface, microphone interface received poor sound quality due to impedance mismatch, which cannot be used)

Connect the serial port of the modem to the serial port of the PC.

Insert the telephone line to the line port of the modem.

Next, turn on the modem power, use securecrt to connect to the serial port, and enter some AT commands to check whether the modem can be executed.

Well, the hardware preparation is complete.

　　

PS: to be continued