Sonic communication - streaming data processing - window alignment

Problem description

• The principle of the acoustic communication program currently implemented by the individual is ： It is expressed by two sound waves of different frequencies 0 and 1, According to the binary code of the data to be transmitted, a sound wave is generated for playing , The audio receiver uses the fast Fourier algorithm to analyze the collected audio data , The binary data is restored by analyzing the arrangement order of two specific frequencies .
• problem ： Audio frequency is not a sample The audio can be parsed out , Use the fast Fourier algorithm to analyze , Multiple required sample Audio data can represent a frequency , Suppose every 40 individual sample Represents a frequency （ It's a binary bit 0 or 1）, The audio receiver also needs to use the same sample Number as a parsing window , Because the time of sending and receiving is not necessarily synchronized , There will be a problem of window alignment , as follows ：

*  Suppose the sending data is 0 and 1 alternate , The window is 4 individual sample Audio data for 
 The sender ：000011110000111100001111000011110000111
 The receiver ：0000111100001111....
 May be ： 000111100001111....
 It could be ：00111100001111....
 It could be ： 0111100001111....

• Pictured above , If the receiving and parsing end also follows 4 individual sample Parse , When the window is not aligned ,4 individual sample It will be mixed in 0 and 1, The parsed data may be wrong , Or it's all a mess of data .

Problem solving

• The solution given by the old employees of the company before is ： Create multiple parsing processes , Each process is separated by a certain time , In this way, there is always a completely correct analysis .
• After a certain period of thinking , There's a better way .

advantage

• When determining the frequency of audio data in a window , The fast Fourier operation value for judging two specific frequencies can be used , Which high is considered to be which frequency , In this case, the window does not need to be fully aligned , As long as the correct frequency in the window exceeds another frequency , That's right sample Data accounts for more than 50% Can parse out the correct data , as follows ：

*  If the data is 0
 Received audio sample data ：0000 （ Can correctly parse ）
0001（ Can correctly parse ）
0011（ Can't parse correctly ）

Real scene

• Above , As long as the parsing window is not exactly half the length of the sending window, there will be no error , The probability of this situation is also relatively small , In this case , In theory, the correct answer can be found in most cases ; But the actual test found that , No other treatment , The probability of resolving the correct result is very low , as a result of ： Both sender and receiver may lose sample.
• Correct solution ： The closer the parsing window is to the start or end of the sending window , The more likely you are to get the right result .

How to achieve

• We cannot guarantee the time difference between a single parsing window and a sending window , So we need to create additional parsing windows , Just create another one , You don't need more than one , The time difference between the first parsing window and the first parsing window is controlled to be half of the window , as follows ：

 The sender ：000011110000111100001111000011110000111
*  situation 1： The first window parsing error , The second window can parse correctly 
 Receiver window 1：00111100001111.... 
 Receiver window 2：  111100001111.... （ It is half a window longer than the first window ）
*  situation 1： The first window is parsed correctly , The second window will parse the error 
 Receiver window 1：000111100001111....
 Receiver window 2：  0111100001111.... （ It is half a window longer than the first window ）