dsp.BinaryFileReader

Read data from binary file

Description

The dsp.BinaryFileReader System object™ reads multichannel signal data from a binary file. If the header is not empty, then the header precedes the signal data. The System object specifies the prototype of the header, and the type, size, and complexity of the data. The first time you read the file, the reader reads the header, followed by the data. On subsequent calls, the reader reads the remaining data. Once the end of file is reached, the reader returns zeros of the specified data type, size, and complexity. The reader can read signal data from a binary file that is not created by the dsp.BinaryFileWriter System object.

The object accepts floating-point data or integer data. To read character data and fixed-point data, see the Write and Read Character Data and Write and Read Fixed-Point Data examples. The input data can be real or complex. When the data is complex, the object reads the data as interleaved real and imaginary components. For an example, see Read Complex Data. The reader assumes the default endianness of the host machine. To change the endianness, you can use the swapbytes function. For an example, see Change Endianness of Data.

To read data from a binary file:

  1. Create the dsp.BinaryFileReader object and set its properties.

  2. Call the object with arguments, as if it were a function.

To learn more about how System objects work, see What Are System Objects? (MATLAB).

Creation

Syntax

reader = dsp.BinaryFileReader
reader = dsp.BinaryFileReader(fname)
reader = dsp.BinaryFileReader(fname,Name,Value)

Description

reader = dsp.BinaryFileReader creates a binary file reader object, reader, using the default properties.

example

reader = dsp.BinaryFileReader(fname) sets the Filename property to fname.

reader = dsp.BinaryFileReader(fname,Name,Value) with Filename set to fname, and each specified property set to the specified value. Unspecified properties have default values.

Example: reader = dsp.BinaryFileReader('myFilename.bin','SamplesPerFrame',1000,'NumChannels',2);

Properties

expand all

Unless otherwise indicated, properties are nontunable, which means you cannot change their values after calling the object. Objects lock when you call them, and the release function unlocks them.

If a property is tunable, you can change its value at any time.

For more information on changing property values, see System Design in MATLAB Using System Objects (MATLAB).

Name of the file from which the object reads the data, specified as a character vector. If the file is not on the MATLAB® path, then specify the full path for the file.

The structure specifies the prototype of the file header, that is, the size of the header and the data type of the field values. The structure can have an arbitrary number of fields. Each field of the structure must be a real matrix of a built-in type. For example, if HeaderStructure is set to struct('field1',1:10,'field2',single(1)), the object assumes that the header is formed by 10 real double-precision values followed by 1 single-precision value. If the file contains no header, you can set this property to an empty structure, struct([]). To retrieve the file header, call the readHeader method on the reader object.

Number of samples per output frame, specified as a positive integer. SamplesPerFrame specifies the number of rows of the output matrix that the object returns. The size of the data is SamplesPerFrame-by-NumChannels. Once the end of file is reached, the reader returns zeros of the specified data type, size, and complexity.

Number of channels, specified as a positive integer. NumChannels specifies the number of columns of the output matrix that the object returns. This property defines the number of consecutive interleaved data samples stored in the file for each time instant. The size of the data is SamplesPerFrame-by-NumChannels. Once the end of file is reached, if the output matrix is not full, the object fills the matrix with zeros to make it a full-sized matrix.

Type of data in file, specified as a character vector. This property defines the data type of the matrix returned by the object algorithm.

Option to specify data complexity, specified as false or true. When this property is set to true, the reader treats the data as complex. The object reads the data as interleaved real and imaginary components. Consider a reader object configured to read the data as a 2-by-2 matrix. The object reads [1 5 2 6 3 7 4 8] as [1 2; 3 4]+1j*[5 6; 7 8]. If this property is set to false, the same object reads the data as [1 5; 2 6].

Usage

Syntax

data = reader()

Description

example

data = reader() reads data from the binary file in a row-major format. The data type, size, and complexity of the data are determined by the properties of the reader object. Once the end of file is reached, the output contains zeros of the specified data type, size, and complexity.

Output Arguments

expand all

Data read by the binary file reader, returned as a vector or a matrix. The size of the data is given by SamplesPerFrame-by-NumChannels, where SamplesPerFrame and NumChannels are the properties of the dsp.BinaryFileReader object.

Object Functions

To use an object function, specify the System object as the first input argument. For example, to release system resources of a System object named obj, use this syntax:

release(obj)

expand all

isDoneEnd-of-data status
readHeaderRead file header
stepRun System object algorithm
releaseRelease resources and allow changes to System object property values and input characteristics
resetReset internal states of System object

Examples

expand all

Create a binary file with a custom header using the dsp.BinaryFileWriter System object. Write data to this file. Read the header and data using the dsp.BinaryFileReader System object.

Write the Data

Specify the file header as a structure with the following fields:

  • DataType set to double.

  • Complexity set to false.

  • FrameSize (number of rows in the data matrix) set to 150.

  • NumChannels (number of columns in the data matrix) set to 1.

Create a dsp.BinaryFileWriter object using this header. The object writes the header first, followed by the data, to ex_file.bin. The data is a noisy sine wave signal. View the data in a time scope.

L = 150;
header = struct('DataType','double','Complexity',false,'FrameSize',L,'NumChannels',1);
writer = dsp.BinaryFileWriter('ex_file.bin','HeaderStructure',header);

sine = dsp.SineWave('SamplesPerFrame',L);
scopewriter = dsp.TimeScope(1,'YLimits',[-1.5 1.5],'SampleRate',...
    sine.SampleRate,'TimeSpan',1);

for i = 1:1000
    data = sine() + 0.01*randn(L,1);
    writer(data);
    scopewriter(data)
end

Release the writer so that the reader can access the data from this file.

release(writer);

Read the Data

Read the data from the binary file, ex_file.bin, using the dsp.BinaryFileReader object. The file contains the header data followed by the actual data. The object reads the binary data until the end of file is reached. Specify the header to the reader using the HeaderStructure property of the reader object.

If the exact header is not known on the reader side, you must at least specify the prototype of the header. That is, the number of fields, and the data type, size, and complexity of each field in the prototype must match with the header data written to the binary file. When the readHeader function reads the data from the binary file, the function extracts the header information based on how the fields are specified in the header prototype. For example, a header field set to 'double' on the writer side can be specified as any string of 6 characters on the reader side. The readHeader function reads this field as a string of 6 characters from the binary file, which matches with 'double'.

headerPrototype = struct('DataType','datype','Complexity',false,'FrameSize',1,'NumChannels',10);
reader = dsp.BinaryFileReader(...
    'ex_file.bin',...
    'HeaderStructure',headerPrototype);
headerReader = readHeader(reader)
headerReader = 

  struct with fields:

       DataType: 'double'
     Complexity: 0
      FrameSize: 150
    NumChannels: 1

The header data extracted by the readHeader function is assigned to the corresponding properties of the reader object.

reader.IsDataComplex = headerReader.Complexity;
reader.DataType = headerReader.DataType;
reader.NumChannels = headerReader.NumChannels;
reader.SamplesPerFrame = headerReader.FrameSize;

Initialize a scope on the reader side to view the extracted binary file data.

scopereader = dsp.TimeScope(1,'YLimits',[-1.5 1.5],'SampleRate',...
    sine.SampleRate,'TimeSpan',1);

The data is read into a single channel (column) containing multiple frames, where each frame has 150 samples. View the data in a time scope.

while ~isDone(reader)
    out = reader();
    scopereader(out)
end
release(reader);
release(scopereader);

Set the reader to read data in frames of size 300. Verify that the data read matches the data written to the file.

reader.SamplesPerFrame = 300;
while ~isDone(reader)
    out = reader();
    scopereader(out)
end
release(reader);

Even when the reader reads data with a different frame size, the output in both time scopes matches exactly.

Use a dsp.BinaryFileReader System object™ to read data from a binary file in a row-major format.

Write the Data

Write the matrix A to the binary file Matdata.bin using a dsp.BinaryFileWriter object. The object writes the specified header followed by the data.

The header has the following format:

  • DataType set to double.

  • Complexity set to false.

  • FrameSize (number of rows in the data matrix) set to 3.

  • NumChannels (number of columns in the data matrix) set to 4.

A = [1 2 3 8; 4 5 6 10; 7 8 9 11];
header = struct('DataType','double','Complexity',false,'FrameSize',3,'NumChannels',4);
writer = dsp.BinaryFileWriter('Matdata.bin','HeaderStructure',header);
writer(A);

Release the writer so that the reader can access the data.

release(writer);

Read the Data

Specify the header using the HeaderStructure property of the reader object. If the exact header is not known, you must at least specify the prototype of the header. That is, the number of fields, and the data type, size, and complexity of each field in the prototype must match with the header data written to the binary file. The dsp.BinaryFileReader object reads the binary file Matdata.bin until the end of file is reached. Configure the System object to read the data into 4 channels, with each channel containing 5 samples. Each loop of the iteration reads a channel (or frame) of data.

headerPrototype = struct('DataType','double','Complexity',false,'FrameSize',5,'NumChannels',4);
reader = dsp.BinaryFileReader('Matdata.bin','HeaderStructure',headerPrototype,'NumChannels',4,...
    'SamplesPerFrame',5);
while ~isDone(reader)
    out = reader();
    display(out)
end
out = 5×4

     1     2     3     8
     4     5     6    10
     7     8     9    11
     0     0     0     0
     0     0     0     0

Each frame of out contains frames of the matrix A, followed by zeros to complete the frame. The original matrix A contains 4 channels with 3 samples in each channel. The reader is configured to read data into 4 channels, with each channel containing 5 samples. Because there are not enough samples to complete the frame, the reader object appends zeros at the end of each frame.

Read the header data from a binary file using the readHeader function.

Write a header, followed by the data to a binary file named myfile.dat. The header is a 1-by-4 matrix of double precision values, followed by a 5-by-1 vector of single-precision values. The data is a sequence of 1000 double-precision values.

fid = fopen('myfile.dat','w');
fwrite(fid,[1 2 3 4],'double');
fwrite(fid,single((1:5).'),'single');
fwrite(fid,(1:1000).','double');
fclose(fid);

Read the header using a dsp.BinaryFileReader object. Specify the expected header structure. This structure specifies only the format of the expected binary file header and does not contain the exact values.

reader = dsp.BinaryFileReader('myfile.dat');
s = struct('A',zeros(1,4),'B',ones(5,1,'single'));
reader.HeaderStructure = s;

Read the header using the readHeader function.

H = readHeader(reader);
fprintf('H.A: ')
H.A: 
fprintf('%d ',H.A);
1 2 3 4 
fprintf('\nH.A datatype: %s\n',class(H.A))
H.A datatype: double
fprintf('H.B: ')
H.B: 
fprintf('%d ',H.B);
1 2 3 4 5 
fprintf('\nH.B datatype: %s\n',class(H.B))
H.B datatype: single

Read complex data from a binary file using the dsp.BinaryFileReader object.

Write a sequence of numbers to a binary file named myfile.dat. There is no header. The data is a 2-by-4 matrix of double-precision values. fwrite writes the data in a column-major format. That is, the 2-by-4 matrix [1 2 3 4; 9 10 11 12] is written as [1 9 2 10 3 11 4 12] in the binary file.

fid = fopen('myfile.dat','w');
fwrite(fid,[1 2 3 4; 9 10 11 12],'double');
fclose(fid);

Specify the data to be complex using the IsDataComplex property. The object reads the data as interleaved real and imaginary components. The SamplesPerFrame and NumChannel properties specify the number of rows and columns of the output data. The header structure is specified as empty.

reader = dsp.BinaryFileReader('myfile.dat','SamplesPerFrame',2,...
    'NumChannels',2,'IsDataComplex',true);
s = struct([]);
reader.HeaderStructure = s;
data = reader();
display(data);
data = 2×2 complex

   1.0000 + 9.0000i   2.0000 +10.0000i
   3.0000 +11.0000i   4.0000 +12.0000i

release(reader);

Alternatively, if you do not specify the data as complex, the reader reads the data as a SamplesPerFrame-by- NumChannel matrix of real values.

reader.IsDataComplex = false;
data = reader();
display(data);
data = 2×2

     1     9
     2    10

release(reader);

The dsp.BinaryFileWriter and dsp.BinaryFileReader System objects do not support writing and reading fixed-point data. As a workaround, you can write the stored integer portion of the fi data, read the data, and use this value to reconstruct the fi data.

Write the Fixed-Point Data

Create an fi object to represent 100 signed random numbers with a word length of 14 and a fraction length of 12. Write the stored integer portion of the fi object to the data file myFile.dat. The built-in data type is int16, which can be computed using class(storeIntData).

data = randn(100,1);
fiDataWriter = fi(data,1,14,12);
storeIntData = storedInteger(fiDataWriter);

writer = dsp.BinaryFileWriter('myFile.dat');
writer(storeIntData);

Release the writer so that the reader can access the data.

release(writer);

Read the Fixed-Point Data

Specify the reader to read the stored integer data as int16 data with 100 samples per data frame. The real-world value of the fixed-point number can be represented using 2(-fractionLength)(storedInteger). If you know the signedness, word length, and fraction length of the fixed-point data, you can reconstruct the fi data using fi(realValue,signedness,wordLength,fractionLength). In this example, the data is signed with a word length of 14 and a fraction length of 12.

reader = dsp.BinaryFileReader('Filename','myFile.dat','SamplesPerFrame',100,...
    'DataType','int16');
data = reader();
fractionLength = 12;
wordLength = 14;
realValue = 2^(-fractionLength)*double(data);

fiDataReader = fi(realValue,1,wordLength,fractionLength);

Verify that the writer data is the same as the reader data.

isequal(fiDataWriter,fiDataReader)
ans = logical
   1

The dsp.BinaryFileWriter and dsp.BinaryFileReader System objects do not support writing and reading characters. As a workaround, cast character data to one of the built-in data types and write the integer data. After the reader reads the data, convert the data to a character using the char function.

Write the Character Data

Cast a character into uint8 using the cast function. Write the cast data to the data file myFile.dat.

data = 'binary_file';
castData = cast(data,'uint8');
writer = dsp.BinaryFileWriter('myFile.dat');
writer(castData);

Release the writer so that the reader can access the data.

release(writer);

Read the uint8 Data

Configure the reader to read the cast data as uint8 data.

reader = dsp.BinaryFileReader('myFile.dat','DataType','uint8','SamplesPerFrame',11);
readerData = reader();
charData = char(readerData);

Verify that the writer data is the same as the reader data. By default, the reader returns the data in a column-major format.

strcmp(data,charData.')
ans = logical
   1

By default, the dsp.BinaryFileReader System object™ uses the endianness of the host machine. To change the endianness, such as when the host machine that writes the data does not have the same endianness as the host machine that reads the data, use the swapbytes function.

Write a numeric array into myfile.dat in big endian format. Read the data using the dsp.BinaryFileReader object. The reader object reads the data in little endian format.

fid = fopen('myfile.dat','w','b');
fwrite(fid,[1 2 3 4 5 6 7 8],'double');
fclose(fid);
reader = dsp.BinaryFileReader('myfile.dat','SamplesPerFrame',8);
x = reader();
display(x);
x = 8×1
10-318 ×

    0.3039
    0.0003
    0.0104
    0.0206
    0.0256
    0.0307
    0.0357
    0.0408

x does not match the original data. Change the endianness of x using the swapbytes function.

y = swapbytes(x);
display(y);
y = 8×1

     1
     2
     3
     4
     5
     6
     7
     8

y matches the original data.

Extended Capabilities

Introduced in R2016b