An ancient topic - About busy sound automatic identification

xiaoxiao2021-03-17  231

In the previous section, Yesongldz tells the general problem about busy tone in a near joke. Its main consciousness is two points:

First, busy sound signal is a communication standard;

Second, the hearing feeling of busy sound signal;

This section mainly describes the machine automatic identification of busy tone signals.

The frequency of the busy tone signal is 450 Hz, with a certain pronunciation and mute interval, the national standard is limited to 350 milliseconds. This particular frequency and interval response characteristic, people listen to "Duda ......................................... Ordinary people are very easy to judge.

If you let the machine to "listen" busy, how can it be judged? Among the real society, there are a lot of terminal switches, board and smart terminal equipment to listen to busy tone signals, how is it implemented? And listen to me slowly.

It can be judged that the busy sound is listening, and the people are listening, and they should know, and the machine is calculated. How is the machine calculated?

First, if it is possible to calculate by the machine, the data is indispensable. The telephone line is transmitted is analog electrical signal, which cannot be directly used as a basis for the calculation. The analog-to-digital conversion circuit is generally employed, and the analog signal is transformed into a value that can be calculated. What are the standard characteristics of these values? Yes, generally use PCM system, A law or μ.

Second, decompress the data. Since the A law or μ is 8-bit compressed data, you cannot directly calculate, convert it to linear data. Filter calculations in the form of linear data, advance rows, and then calculate energy in accordance with a certain formula.

Again, the calculated energy is compared to the preset energy threshold of the standard signal, and the signal sound space is compared, and finally confirmed whether or not it is busy.

PCM system can be understood to be 8,000 data per second per second. For each data, then filter, then calculate the energy. The entire busy tone is required for trip to do hundreds of thousands of calculations per second, and the performance requirements of the microprocessor are high.

We know that it can judge busy tone, it is to pass audio, form a sense of hearing, and then judge by the brain. Here, the ear is equivalent to a filter, a brain shape such as a supercomputer. As mentioned in the foreman, the hearing range range between 20Hz and 20000 Hz. That is, the human ear is similar to a band pass filter for 20 h to 20Hz, less than 20 Hz and more than 20000 Hz audio communication is not accepted by the brain.

YESONGLDZ is not intended here to detail the specific algorithm. Only a few points are introduced.

Point 1: The calculation frequency range is selected from 300 to 700 Hz.

Have you used to say that the frequency of the standard Huaai signal is 450 Hz, what is calculated so wide? Standards and actual, just like the theory and practice, there is no deviation! And not saying that the signal is interfered during the transmission process, saying the exchange of busy tone signals. The service switch cannot guarantee 100% accurate, there are 5% to 10% frequency offset. As for the general small switch, company or community switch, then don't have to say more, the frequency deviation has not specified! I have more than 350Hz and 600 Hz. I have seen more.

IT 2: Calculated signal tone and mute interval between 230 ~ 450 milliseconds

The truth is the same as the upper strip. Standard busy signal 350 millisecond sound, 350 millisecond mute. Some small switches are sent to the sound signal to be 230 milliseconds and mute, some close to 500 milliseconds. More infusable, there are still some small switches, busy tone signals and mute signals are inconsistent, such as busy tone 300 milliseconds, mute 250 milliseconds, of course, this situation is relatively rare.

What should I do with exchange devices that don't match the market in the market? You can't make an automatic identification device that cannot be used in theory and practical practice! In the face of reality, I have to enlarge these two indicators. It is enlarged to the technician's own point that I can't stand it. I have to bite this! Below we will list this, let you see helplessness in some practices. Helpless: misjudgment

The frequency band of the busy audio signal is precisely the frequency band in which the human vocal signal energy is concentrated. People may also make a busy frequency, in theory, there will be misjudgment. In fact, because of a single frequency, there is a unique rhythm. However, from statistics, this kind of chance is small unless a specially trained person speaks to the phone "beep ... beep ...".

Increased frequency polarization range, increase the interval time value of vocal and mute, equal to the default more frequencies and more rhythm, increasing the chance of misjudgment. In fact, this misjury is a large amount from a single indicator, often appearing during calculation. Some busy tone judges that the equipment is very good, and it has made hands and feet from statistics. Then the statistics also brought a less friendly consequence - lag, this is the second helplessness I have to say.

Helpless two: lag

Calculate a busy tone called accident, calculating two still dared to determine, and calculate three only a little grasp. And must be within a relatively short period of time, for example, within, for example, 3 to 5 seconds. Drop and re-statistrate more than this time. Some busy tone recognition equipment actually dare to determine from 12 busy tones. I want to think about it, it is also impossible, who makes there are so many switches that don't match the national standard, you can't always change the customer, you are not expected to have any noise interference on the line!

Helpless three: Record busy tone

Many automatic response devices have a recording function to leave the customer's voice information. Some is timing, let the other party finisken by a set time range, if it is not finished until the limited time is not finished, the rude hangs. There are some smart answering devices that judge the busy tone signal to hang up. Because the device is judged that the busy tone signal takes a short time of a few seconds to more seconds, each customer's message audio file contains a lot of useless busy tone signals, which takes up a lot of storage space.

In order to avoid this problem, advanced recording equipment has a reputable function. After the data is recorded, the audio file is calculated once, and the extra busy tone is filtered out. How is the specific effect, and also read the level of experience in the technician and the luck of the construction site!

What is the "self-learning" intelligent judgment?

Sometimes, we will hear some propaganda, "busy audio signal intelligence judgment" has also added "self-learning" concept. In fact, the so-called "self-learning" is nothing more than the signal response parameters that can be adjusted according to the signal characteristics of the construction site. These busy tone signals do not die related algorithms, but let you and the construction personnel judge according to the site, fill in the appropriate value. This effect will naturally be better.

Ok, I will say this for busy tone signals. Some people will say that there is so many shortcomings in the busy tone signal, why is it still extended to today? That is because of the time to formulate signal standards, who will think of a CTI industry decades, who will think of smart products that will have such a complex application!

转载请注明原文地址:https://www.9cbs.com/read-129560.html

New Post(0)