Advanced Signals and Systems –

Hilbert Transform

Gerhard Schmidt
Christian-Albrechts-Universität zu Kiel
Faculty of Engineering
Institute of Electrical and Information Engineering
Digital Signal Processing and System Theory
Digital Signal Processing and System Theory| Advanced Signals and Systems| Hilbert Transfrom
Contents of the Lecture

Entire Semester:

 Introduction
 Discrete signals and random processes
 Spectra
 Discrete systems
 Idealized linear, shift-invariant systems
 Hilbert transform
 State-space description and system realizations
 Generalizations for signals, systems, and spectra

Digital Signal Processing and System Theory| Advanced Signals and Systems| Hilbert Transform
Transfrom Slide VI-2
Contents of this Part

Hilbert Transform

 Frequency and impulse response of a “Hilbert transformer”

 Frequency-domain definition
 Impulse response
 Hilbert transform, one-sided spectra, and analytic signals
 Definitions
 Example
 Instantaneous amplitude, phase, and frequency
 One-sided signals and causality

Digital Signal Processing and System Theory| Advanced Signals and Systems| Hilbert Transform
Transfrom Slide VI-3
Hilbert Transform

Frequency and Impulse Response of a “Hilbert Transformer” – Part 1

Frequency-domain definition:

A Hilbert transformer is a special case of an ideal, linear-phase system. Such a system is used in
several applications (e.g. modulation theory). The filter is defined by its frequency response:

with

Diagram:

Digital Signal Processing and System Theory| Advanced Signals and Systems| Hilbert Transform
Transfrom Slide VI-4
Hilbert Transform

Frequency and Impulse Response of a “Hilbert Transformer” – Part 2

Frequency-domain definition (continued):

We obtain for the magnitude of the frequency response of a Hilbert filter:

We can see that describes a nearly linear-phase all-pass filter. However, due to the
„jumps“ at the filter is not free of distortions.

The filter achieves a constant phase shift of ±90° with phase jumps at by
180° (in addition to a constant delay of samples).

Digital Signal Processing and System Theory| Advanced Signals and Systems| Hilbert Transform
Transfrom Slide VI-5
Hilbert Transform

Frequency and Impulse Response of a “Hilbert Transformer” – Part 3

Frequency-domain definition (continued):

By applying an inverse Fourier transform we obtain the impulse response of the filter:

… derivation on the blackboard …

Digital Signal Processing and System Theory| Advanced Signals and Systems| Hilbert Transform
Transfrom Slide VI-6
Hilbert Transform

Hilbert Transform, One-sided Spectra, and Analytic Signals – Part 1

Definitions:

The meaning of the filter becomes more obvious if we look at the output of such a
filter (please have a look also on the next slide):

If we add now the input signal and the filtered input signal (after compensation for the filter‘s
delay) we obtain the so-called analytic signal:

Please note, that instead of using a „negative“ delay for the filtered signal a „positive“ delay
can be applied to the input signal!

Digital Signal Processing and System Theory| Advanced Signals and Systems| Hilbert Transform
Transfrom Slide VI-7
Hilbert Transform

Hilbert Transform, One-sided Spectra, and Analytic Signals – Part 2

Definitions (continued):

For the spectrum of an analytic signal we get:

… periodically repeated …

The analytic signal has a one-sided spectrum! This is very useful for a variety of applications.

Digital Signal Processing and System Theory| Advanced Signals and Systems| Hilbert Transform
Transfrom Slide VI-8
Hilbert Transform

Hilbert Transform, One-sided Spectra, and Analytic Signals – Part 3

Definitions (continued):

The signal that is required for the one-sided spectral compensation is called
Hilbert transform . For this transformation the delay is usually set to zero ( )
and a filter with a non-causal impulse response is used:

This relation can be inverted easily (see next slide) and we obtain:

Digital Signal Processing and System Theory| Advanced Signals and Systems| Hilbert Transform
Transfrom Slide VI-9
Hilbert Transform

Hilbert Transform, One-sided Spectra, and Analytic Signals – Part 4

Definitions (continued):

To understand the inversion of the Hilbert transform we start with the application of the Hilbert
filter in the frequency domain. The following lines are restricted to the frequency range
– outside this range periodical expansion is assumed.

… exchanging both sides and dividing by the term in brackets with the sign function …

… expanding the numerator with 1 ...

… truncating the negative imaginary unit „-j“ ...

… exploiting that a multiplication with a sign function is equal to a division with it (except
at 0) ...

Digital Signal Processing and System Theory| Advanced Signals and Systems| Hilbert Transform
Transfrom Slide VI-10
Hilbert Transform

Hilbert Transform, One-sided Spectra, and Analytic Signals – Part 5

Definitions (continued):

Except for the sign we obtain the same relation as the original Hilbert transform. Thus, applying
a Hilbert transform twice leads to the original signal multiplied with -1.

Remarks:
 If we compute the analytic signal of a real input , we will obtain a complex
sequence . The sequences and are
called a pair of Hilbert signals.
 Since for real sequences all information is in the „left“ as well as in the „right“ part
of the spectrum

the analytic signal is uniquely defined by the real input sequence .

Digital Signal Processing and System Theory| Advanced Signals and Systems| Hilbert Transform
Transfrom Slide VI-11
Hilbert Transform

Hilbert Transform, One-sided Spectra, and Analytic Signals – Part 6

Example:

In order to show some applications of the Hilbert transform we will mention now so-called
public address systems as a first example.

In such systems the signal of a speaking

person is recorded by means of a
microphone. After amplification the
Loudspeaker
signal is played back via one or more Amplifier
loudspeakers. This allows for a
better signal-to-noise ratio for Speaking
person
the listeners. However, since Feed-
the loudspeaker signals might back
paths
couple back into the microphone
a closed electro-acoustic loop
is generated.
Microphone

Digital Signal Processing and System Theory| Advanced Signals and Systems| Hilbert Transform
Transfrom Slide VI-12
Hilbert Transform

Hilbert Transform, One-sided Spectra, and Analytic Signals – Part 7

Example:

The transmission from the

loudspeaker over the room to the
microphone determines the maximum
gain that can be used without
generating a non-stable system
(howling).

Countermeasures and
improvements:

 Equalization filters
 Frequency shift filters
(„Hilbert transformer“)

Digital Signal Processing and System Theory| Advanced Signals and Systems| Hilbert Transform
Transfrom Slide VI-13
Hilbert Transform

Hilbert Transform, One-sided Spectra, and Analytic Signals – Part 8

Example (continued):

By means of an appropriate equalization usually a few decibels more gain can be achieved.
However, also a frequency shift is able to increase the maximum gain.

To realize the frequency shift, we

can compute first the analytic signal,
second shift this signal by a few Hertz
Loudspeaker
(about 3 to 10 Hz, realized by means Amplifier
of a modulation) and finally compute
the shifted output signal by means Speaking
person
of a second Hilbert transform. Feed-
back
paths

Equalization and
frequency shift Microphone

Digital Signal Processing and System Theory| Advanced Signals and Systems| Hilbert Transform
Transfrom Slide VI-14
Hilbert Transform

Hilbert Transform, One-sided Spectra, and Analytic Signals – Part 9

Example (continued):

Source: „The Big Bang Theory“, YouTube

Digital Signal Processing and System Theory| Advanced Signals and Systems| Hilbert Transform
Transfrom Slide VI-15
Hilbert Transform

Hilbert Transform, One-sided Spectra, and Analytic Signals – Part 10

Example (continued):

Results:

A gain increase of about 1 to 6 dB

can be achieved (dependent on the
type and size of the room).

If larger shift frequencies would be

applied also larger gains can be
achieved. However, the signal sounds
not very good any more …

0 Hz 5 Hz 10 Hz 15 Hz

20 Hz 25 Hz 30 Hz 35 Hz

Digital Signal Processing and System Theory| Advanced Signals and Systems| Hilbert Transform
Transfrom Slide VI-16
Hilbert Transform

Hilbert Transform, One-sided Spectra, and Analytic Signals – Part 11

Instantaneous amplitude, instantaneous phase, and instantaneous frequency

By means of a Hilbert transform we can change a real input signal to a complex analytic
signal (assuming a delay-free Hilbert filter) and we obtain:

In addition to that we know a well-known form for describing complex quantities:

We can obtain the two quantities from above and compute . This term is called the
instantaneous amplitude or the envelope of :

In a similar way we can compute the instantaneous phase

Digital Signal Processing and System Theory| Advanced Signals and Systems| Hilbert Transform
Transfrom Slide VI-17
Hilbert Transform

Hilbert Transform, One-sided Spectra, and Analytic Signals – Part 12

Instantaneous amplitude, instantaneous phase, and instantaneous frequency (continued)

Similar to the continuous case, where is termed the instantaneous frequency

we define the instantaneous frequency of a discrete signal as

Digital Signal Processing and System Theory| Advanced Signals and Systems| Hilbert Transform
Transfrom Slide VI-18
Hilbert Transform

One-sided Signals and Causality – Part 1

Inverse relations

In the following we will assume a so-called right-hand sided signal , meaning that we
have

Such a signal can be decomposed into an even and an odd part:

Digital Signal Processing and System Theory| Advanced Signals and Systems| Hilbert Transform
Transfrom Slide VI-19
Hilbert Transform

One-sided Signals and Causality – Part 2

Inverse relations (continued)

Please add in the following

diagram the even and the
odd part of the signal!

As we can see, we get the

following relation between
the even and the odd
signal component
(under our assumptions):

Digital Signal Processing and System Theory| Advanced Signals and Systems| Hilbert Transform
Transfrom Slide VI-20
Hilbert Transform

One-sided Signals and Causality – Part 3

Inverse relations (continued)

As a result we can conclude that either or describe the signal completely

(except for the single value ). As a result also either the spectrum

or

fully describe the spectrum . If we transform the relation

into the frequency (Fourier) domain, we obtain

Digital Signal Processing and System Theory| Advanced Signals and Systems| Hilbert Transform
Transfrom Slide VI-21
Hilbert Transform

One-sided Signals and Causality – Part 4

Inverse relations (continued)

This integral is also denoted as a Hilbert transform (even if slightly differently defined in
comparison to its time-domain counterpart):

In addition, we get

Digital Signal Processing and System Theory| Advanced Signals and Systems| Hilbert Transform
Transfrom Slide VI-22
Hilbert Transform

One-sided Signals and Causality – Part 5

Inverse relations (continued)

If the right-hand sided signal is real (in addition) we obtain the following relations for the
spectra

Thus, we obtain for the spectra of real, right-hand sided signals:

The real and the imaginary part of the spectrum

of real, right-hand sided signals are a “pair” of Hilbert transforms

Digital Signal Processing and System Theory| Advanced Signals and Systems| Hilbert Transform
Transfrom Slide VI-23
Hilbert Transform

Contents of the Part on Hilbert Transforms

This part:
 Frequency and impulse response of a “Hilbert transformer”
 Frequency-domain definition
 Impulse response
 Hilbert transform, one-sided spectra, and analytic signals
 Definitions
 Example
 Instantaneous amplitude, phase, and frequency
 One-sided signals and causality

Next part:

 State-space description and system realizations

Digital Signal Processing and System Theory| Advanced Signals and Systems| Hilbert Transform
Transfrom Slide VI-24