(1)

A deterministic signal is a signal that follows a fixed pattern and can be completely described by a
mathematical function. If you know the function that generates a deterministic signal, you can
completely predict its behavior at any point in time. An example of a deterministic signal is a sine
wave, which can be described by the mathematical function y = sin(x).

A stochastic signal, on the other hand, is a signal that is random and cannot be completely described
by a mathematical function. Stochastic signals are typically characterized by a probability distribution
function, which describes the probability of different possible outcomes. An example of a stochastic
signal is white noise, which is a random signal with a uniform probability distribution.

In summary, deterministic signals can be described by a mathematical function, and their future
behavior can be predicted, whereas stochastic signals cannot be described by a mathematical
function and their future behavior is uncertain and follows a probability distribution.

Stochastic signals are typically represented using the concept of a random process. A random
process is a collection of random variables, each of which represents the value of the signal at a
particular point in time. The values of the random variables are not known in advance and are
subject to uncertainty.

One common way to represent a stochastic signal is through its probability distribution function
(PDF). A PDF is a mathematical function that describes the probability of different possible outcomes
for a random variable. For example, the PDF of a normal distribution (also known as a Gaussian
distribution) is a bell-shaped curve, with the mean of the distribution as the peak and the standard
deviation as a measure of the spread of the distribution.

Another way to represent a stochastic signal is using its power spectral density (PSD) function. The
PSD is a representation of the frequency content of a signal and can be calculated using a technique
called Fourier analysis. The PSD can be used to determine the relative strength of the different
frequency components of a signal, and it can be useful in applications like noise reduction and signal
filtering.

Another common method is through the use of autocorrelation and cross-correlation functions
which can be used to represent the statistical properties of the signal such as its temporal
dependencies and periodicity, it can be also useful in applications like filtering, detection,
identification and prediction.

In summary, Stochastic signals can be represented through various ways, depending on the
application and the information that is desired to extract, the most common are probability
distribution function, power spectral density and autocorrelation/cross-correlation functions.

(2)

I)

When a stationary process of order n is applied to stochastic signals, it means that the process is
able to capture the underlying patterns and trends in the signal up to a certain degree of complexity,
represented by the value of n. This means that the process is able to identify patterns in the signal
that repeat themselves over time, such as a trend or a seasonality, and can predict future values
based on these patterns. However, if the signal contains patterns that are more complex than those
captured by the stationary process of order n, the predictions may not be as accurate. Additionally, if
the signal is non-stationary, meaning that the patterns in the signal change over time, the stationary
process may not be able to accurately capture and predict these changes.

ii)

refers to signals that have a constant statistical properties over time, meaning that their mean,
variance, and any other statistical properties do not change over time. This is in contrast to non-
stationary signals, which have time-varying statistical properties. In practice, however, many signals
are only approximately stationary, meaning that they have some small level of variation in their
statistical properties over time. It is important to note that stationary signals in the strict sense are a
mathematical concept and may not always reflect the reality of real-world signals.

iii)

Wide sense stationary (WSS) is a term used to describe a type of stochastic signal that exhibits
certain statistical properties over time. Specifically, a WSS signal is one where the mean and variance
of the signal are constant over time, and the correlation between any two points in the signal is only
a function of the time difference between them.

This means that if we were to take a series of samples of the signal at different points in time, the
mean and variance of those samples would be the same, and the correlation between any two
samples would only depend on how far apart they were taken in time.

WSS signals are useful in signal processing and communications because they can be modeled and
analyzed using a variety of techniques, such as Fourier analysis or filtering. Additionally, WSS signals
are often used as a model for real-world signals, such as temperature or noise, that exhibit similar
statistical properties.

However, it is important to note that not all signals are WSS. Some signals, such as speech or music,
exhibit changing statistics over time and would not be considered WSS. In these cases, more
advanced signal processing techniques are required to analyze and model the signal.

(3)

The discrete autocorrelation function (DACF) is a statistical measure that describes the correlation
between a signal and a time-shifted version of itself. It is used to quantify the similarity between a
signal and its delayed versions, and it is often used to study the properties of a signal and how it
changes over time.

The DACF is defined mathematically as the cross-correlation of a signal x(n) with a time-shifted
version of itself, x(n-k), where k is the time delay. The equation for the DACF is:

Rxx(k) = 1/N * Sum(x(n) * x(n-k)) for n = 0 to N-1

Where Rxx(k) is the DACF, N is the length of the signal, and the sum is taken over all values of n from
0 to N-1. The result of the equation is a value for each possible time delay, k, that describes the
correlation between the signal and its delayed version.

The DACF can be used to extract information about the frequency components of a signal, the
stationarity and persistence of a signal, and the presence of patterns or periodicities in a signal. It is
also used to determine the order of linear systems, identify the presence of non-stationarity in a
signal, and in some filtering applications

PROPERTIES OF AUTOCORRELATION

The discrete autocorrelation function (ACF) is a mathematical function that describes the correlation
between different samples of a discrete-time signal. The properties of the ACF include:

1. Symmetry: The ACF is symmetric about the origin, meaning that the correlation between
samples x[n] and x[n-k] is the same as the correlation between samples x[-n] and x[-n+k].

2. Normalization: The ACF is usually normalized such that its value at the zero-lag (i.e., when k
= 0) is equal to 1. This is known as the autocorrelation coefficient.

3. Energy conservation: The integral or sum of the autocorrelation function is equal to the
energy of the signal.

4. Stationarity: If the signal is wide-sense stationary, the ACF will only depend on the time
difference between the samples, not on the specific time or location.

5. Real-valued: The ACF is a real-valued function, meaning it only takes on real values, not
complex values.

6. Non-decreasing: The ACF is non-decreasing in the absolute sense, i.e., the correlation
between a signal and a delayed version of itself will never be negative.

7. Zero outside the signal: The autocorrelation function is zero for lags greater than the signal
length.

These properties of the autocorrelation function can be used to analyze a signal, such as identifying
its periodicity, detecting the presence of noise, and detecting non-stationarity in the signal.

Relation to the Power Spectral Density (PSD) of a Signal

The discrete autocorrelation function (ACF) and the Power Spectral Density (PSD) are related
because they both provide information about the statistical properties of a signal. However, they
provide this information in different ways.

The ACF measures the correlation between different samples of the signal. It tells us how similar
different parts of the signal are to each other. The PSD, on the other hand, measures the power of
the different frequency components of the signal. It tells us how much energy is present at different
frequencies.
The relationship between the ACF and the PSD is given by the Wiener-Khinchin theorem, which
states that the PSD of a wide-sense stationary signal is the Fourier transform of its autocorrelation
function. This means that the PSD and the ACF can be used to determine each other.

For example, if we know the PSD of a signal, we can use the inverse Fourier transform to determine
its ACF. Similarly, if we know the ACF of a signal, we can use the Fourier transform to determine its
PSD.

It's important to note that the relationship between the ACF and the PSD holds only for wide-sense
stationary signals, since the Wiener-Khinchin theorem relies on the stationarity assumption. Also,
the relationship works only for discrete time signals, the continuous case would be a different story.

SECTION B;

1)

There are several different architectures that can be used for signal processing, each with its own
advantages and disadvantages. Some of the most common architectures include:

1. Von Neumann architecture: This is the traditional architecture that is used in most modern
computers. It is based on the stored program concept, where a program is stored in memory
and executed by a central processing unit (CPU). The Von Neumann architecture is well-
suited for tasks that involve a lot of branching and data manipulation, but it is less efficient
for tasks that involve large amounts of data and parallel processing.

2. Flynn's taxonomy: This is a classification system for computer architectures based on the
number of concurrent instruction streams and data streams. It divides architectures into
four categories: Single instruction, single data (SISD), Single instruction, multiple data
(SIMD), multiple instruction, single data (MISD), and multiple instruction, multiple data
(MIMD). These architectures are typically used in various specific-purpose applications.

3. DSP (Digital Signal Processing) processors: these are specialized processors that are designed
specifically for digital signal processing tasks. DSP processors have features such as high-
performance fixed-point arithmetic and specialized instructions for common signal
processing operations, such as fast Fourier transforms (FFT) and digital filtering. These
processors are well-suited for tasks such as audio and video processing, wireless
communication, and control systems.

4. FPGA (Field-Programmable Gate Array) based architectures: FPGA-based architectures use

programmable logic devices (PLDs) to implement digital systems. These architectures are
highly flexible and can be reprogrammed to perform different tasks, making them well-
suited for tasks that involve a lot of parallel processing. They can be also used as co-
processors to Von Neumann architectures to speedup specific tasks.

5. GPU (Graphics Processing Unit) based architectures: These architectures use specialized
processors that are designed to handle large amounts of data and perform parallel
processing quickly. These processors are primarily used for tasks related to computer
graphics, but they are also well-suited for other tasks such as deep learning and scientific
computing.
These are some of the most common architectures that can be used for signal processing, but there
are many other options as well. The choice of architecture will depend on the specific requirements
of the signal processing task, and the trade-offs between cost, performance, and flexibility.