Digital signal processing ( DSP ) is the use of digital processing, such as by computers or more specialized digital signal processors, to perform a variety of signal processing operations. Signals processed in this way are sequence numbers that represent samples of continuous variables in domains such as time, space, or frequency.
Digital signal processing and analog signal processing are subfields of signal processing. DSP applications include audio and speech processing, sonar, radar and other sensor array processing, spectral density estimates, statistical signal processing, digital image processing, signal processing for telecommunications, control systems, biomedical engineering, seismology, and so on.
DSP can involve linear or nonlinear operation. Nonlinear signal processing is closely related to the identification of nonlinear systems and can be implemented in time, frequency, and spatio-temporal domains.
Digital computing applications for signal processing allow many advantages over analog processing in many applications, such as error detection and correction in transmission and data compression. DSP applies to data streaming and static data (stored).
Video Digital signal processing
Sampling signal
To digitally analyze and manipulate analog signals, it must be digitized with analog-to-digital converter (ADC). Sampling is usually done in two stages, discretization and quantization. Discreteisation means that the signal is divided into the same time interval, and each interval is represented by one amplitude measurement. Quantization means that each amplitude measurement is approximated by the value of a finite set. Rounding the real number to an integer is an example.
The Nyquist-Shannon sampling theorem states that the signal can be precisely reconstructed from the sample if the sampling frequency is greater than twice the highest frequency component in the signal. In practice, sampling frequencies are often significantly higher than twice the Nyquist frequency.
Theoretical DSP and derivation analysis is usually performed on time-discrete signal models without the amplitude inaccuracies (quantization errors), "made" by the abstract sampling process. Numerical methods require quantized signals, such as those generated by ADCs. Processed results may be a frequency spectrum or a set of statistics. But often this is another quantized signal that is converted back to analog form by a digital-to-analog converter (DAC).
Maps Digital signal processing
Domains
In DSP, engineers typically study digital signals in one of the following domains: time domain (one-dimensional signal), spatial domain (multidimensional signal), frequency domain, and wavelet domain. They choose domains to process signals by making informed assumptions (or by trying different possibilities) to determine which domain best represents the important characteristics of the signal and the processing that will be applied to it. The sample sequence of the measuring device yields a temporal or spatial domain representation, whereas the discrete Fourier transform produces a frequency domain representation.
Time and space domains
The most common processing approach in the time or space domain is the increase of the input signal through a method called filtering. Digital filtering generally consists of several linear transformations of a number of samples around the sample of the current from the input or output signals. There are different ways to characterize filters; as an example:
- The linear filter is a linear transformation of the input sample; the other filter is nonlinear . Linear filters meet the principle of superposition, that is, if the input is a weighted linear combination of different signals, the output is the same weighted linear combination of the corresponding output signal.
- The causal filter uses only the previous sample of the input or output signals; while a non-causal filter uses a sample input in the future. A non-causal filter can usually be converted into a causal filter by adding a delay to it.
- The time-invariant filter has constant properties over time; Other filters like adaptive filters change over time.
- A stable filter produces a unified output to a constant value with time, or remains limited in limited intervals. Unstable filters can produce unlimited outputs, with restricted or even zero inputs.
- An impulse response filter (FIR) is limited to input signals only, while an infinite response impulse filter (IIR) uses an input signal and a previous sample of the output signal. FIR filters are always stable, while IIR filters may be unstable.
Filters can be represented by block diagrams, which can then be used to derive sample processing algorithms to apply filters with hardware instructions. Filters can also be described as a difference equation, a set of zero and poles or an impulse response or step response.
The output from a linear digital filter to a given input can be calculated by converting the input signal with the impulse response.
Frequency domains
Signals are converted from time or space domains to frequency domain usually through the use of Fourier transform. The Fourier transform alters the time or space information into the magnitude and phase components of each frequency. With some applications, how the phase varies with frequency can be a significant consideration. Where the phase is not important, the Fourier transform is often converted to the power spectrum, which is the magnitude of each quadratic frequency component.
The most common goal for signal analysis in frequency domain is signal property analysis. Engineers can study the spectrum to determine which frequencies are in the input signal and which ones are missing. Frequency domain analysis is also called spectrum - or spectral analysis .
Filtering, especially in non-real-time jobs can also be reached within the frequency domain, applying filters and then converting back to time domain. This can be an efficient implementation and can basically provide filter responses including an excellent approximation to the brickwall filter.
There are several frequency domain transformations that are often used. For example, cepstrum converts the signal to the frequency domain through the Fourier transform, takes the logarithm, then implements another Fourier transform. It emphasizes the harmonic structure of the original spectrum.
Analysis of Z-plane
Digital filters are available in IIR and FIR types. FIR filters have many advantages, but are computationally more demanding. While the FIR filter is always stable, the IIR filter has a feedback loop that may resonate when stimulated with a particular input signal. Z-transform provides a tool to analyze the potential of digital IIR filter stability problem. This is analogous to the Laplace transform, which is used to design analog IIR filters.
Wavelet
In numerical analysis and functional analysis, discrete wavelet transforms (DWT) are any wavelet transforms whose wavelets are randomly selected. Like other wavelet transformations, the main advantage it has on the Fourier transform is its temporal resolution: capturing both the and frequency of location information. The accuracy of the shared time-frequency resolution is limited by the time-frequency uncertainty principle.
Apps
The main applications of DSP are audio signal processing, audio compression, digital image processing, video compression, speech processing, speech recognition, digital communication, digital synthesizer, radar, sonar, financial signal processing, seismology and biomedical. Specific examples are compression and voice transmission in digital cell phones, sound room correction in hi-fi and sound amplification applications, weather forecasts, economic forecasts, seismic data processing, industrial process control and analysis, medical imaging such as CAT scan and MRI, MP3 compression , computer graphics, image manipulation, hi-fi loudspeaker crossover and equalization, and audio effects for use with electric guitar amplifiers.
Implementation
The DSP algorithm has long been run on general purpose computers and digital signal processors. The DSP algorithm is also implemented on custom-made hardware such as application-specific integrated circuits (ASIC). Additional technologies for digital signal processing include more powerful general-purpose microprocessors, field-programmable gate interfaces (FPGAs), digital signal controllers (mostly for industrial applications such as motor controls), and flow processors.
Depending on the requirements of the application, digital signal processing tasks can be implemented on a general purpose computer.
Often when the processing requirement is not real-time, processing is economically done with existing general-purpose computers and signal data (input or output) is in the data file. This is basically no different from other data processing, except DSP math techniques (such as FFT) are used, and sample data are usually assumed to be uniformly sampled in time or space. For example: processing digital photos with software like Photoshop .
However, when the application needs are real-time, DSP is often implemented using special microprocessors such as DSP56000, TMS320, or SHARC. It often processes data using fixed-point arithmetic, although some stronger versions use floating point. For faster applications, FPGA may be used. Beginning in 2007, multicore implementations of DSP have started to emerge from companies including Freescale and Stream Processors, Inc. For faster applications with broad usage, ASIC may be specially designed. For slow applications, slower traditional processors such as microcontrollers may be adequate. Also a growing number of DSP applications are now implemented on embedded systems using powerful PCs with multi-core processors.
Technique
Related fields
Note
References
Further reading
Source of the article : Wikipedia
0 Comments