Short-Time FFT

Nonparametric estimate of spectrum using short-time, fast Fourier transform (FFT) method

Library

Transforms

dspxfrm3

Description

The Short-Time FFT block computes a nonparametric estimate of the spectrum. The block buffers, applies a window, and zero pads the input signal. The block then takes the FFT of the signal, transforming it into the frequency domain.

Connect your single-channel analysis window to the w(n) port. For the Analysis window length parameter, enter the length of the analysis window, W. The block buffers the input signal such that it has a frame length of W

Connect your single-channel or multichannel input signal to the x(n) port. After the block buffers and windows this signal, it zero-pads the signal before computing the FFT. For the FFT length parameter, enter the length to which the block pads the input signal. For the Overlap between consecutive windows (in samples) parameter, enter the number of samples to overlap each frame of the input signal.

The block outputs the complex-valued, single-channel or multichannel short-time FFT at port X(n,k).

Fixed-Point Data Types

The following diagram shows the data types used within the Short-Time FFT subsystem block for fixed-point signals.

The settings for the fixed-point parameters of the Array-Vector Multiply block in the diagram above are as follows:

• Rounding Mode — Floor

• Saturate on integer overflow — Wrap

• Product output — Inherit via internal rule

• Accumulator — Inherit via internal rule

• Output — Same as first input

The settings for the fixed-point parameters of the FFT block in the diagram above are as follows:

• Rounding Mode — Floor

• Saturate on integer overflow — Wrap

• Sine table — Same word length as input

• Product output — Inherit via internal rule

• Accumulator — Inherit via internal rule

• Output — Inherit via internal rule

See the FFT and Array-Vector Multiply block reference pages for more information.

Examples

The dspstsa example illustrates how to use the Short-Time FFT and Inverse Short-Time FFT blocks to remove the background noise from a speech signal. To open the dspstsa model, type dspstsa in the MATLAB^® command prompt.

Parameters

Analysis window length

Specify the frame length of the analysis window. The Analysis window length must be a positive integer value greater than one.

Overlap between consecutive windows (in samples)

Enter the number of samples of overlap for each frame of the input signal.

FFT length

Enter the length to which the block pads the input signal.

Supported Data Types

Port	Supported Data Types
x(n)	Double-precision floating point Single-precision floating point Fixed point (signed only) 8-, 16-, and 32-bit signed integers
w(n)	Double-precision floating point Single-precision floating point Fixed point (signed only) 8-, 16-, and 32-bit signed integers
X(n,k)	Double-precision floating point Single-precision floating point Fixed point (signed only) 8-, 16-, and 32-bit signed integers

References

[1] Quatieri, Thomas E. Discrete-Time Speech Signal Processing. Englewood Cliffs, NJ: Prentice-Hall, 2001.

Extended Capabilities

C/C++ Code Generation

Generate C and C++ code using Simulink® Coder™.

Usage notes and limitations:

• Generated code relies on memcpy or memset functions (string.h) under certain conditions.

• When the FFT length is not a power of two, the executable generated from this block relies on prebuilt dynamic library files (.dll files) included with MATLAB. Use the packNGo function to package the code generated from this block and all the relevant files in a compressed zip file. Using this zip file, you can relocate, unpack, and rebuild your project in another development environment where MATLAB is not installed. For more details, see How To Run a Generated Executable Outside MATLAB.

  When the FFT length is a power of two, you can generate standalone C and C++ code from this block.

Fixed-Point Conversion

Convert floating-point algorithms to fixed point using Fixed-Point Designer™.

If the input is fixed point, it must be signed integer or signed fixed point with power-of-two slope and zero bias.