Concentración en

Neurociencias Cognitivas

Digitalización de Señales
Dr. Mauricio Ramírez
mauricio.ramirezm@tec.mx
Fourier Series 𝑥(𝑡) = 𝐴 ∗ 𝑠𝑖𝑛 2𝜋𝑓𝑡 + θ = 𝐴 ∗ sin 𝑤𝑝 𝑡 + θ = 𝐴 ∗ sin(
2𝜋𝑡
)
𝑇

Any periodic signal can be Decomposed signals (by frequency)

decomposed into a sum of
multiple waveforms with
independent properties.

Each waveform has their own

• Amplitude (A)
• Frequency (f)
• Phase (𝜑, or θ) 1
𝑇=
Original signal 𝑓
Transformada de Fourier
Expresión matemática que se utiliza para pasar una señal del dominio del tiempo
al dominio de la frecuencia.

𝑁−1

𝑋 𝑘 = ෍ 𝑥(𝑛)𝑒 −𝑗2𝜋𝑘𝑛/𝑁
𝑛=0

𝑋 𝑘 - real (𝑓) e imaginario (𝜑)

𝑥(𝑛) – valores discretos de la señal

𝑁 – longitud de la señal

𝑘 – dominio frecuencial 𝑘 = 1, … , 𝑁 − 1
Types of Signals

• Continuous (analog).
Continuous samples across
all time points.

• Discrete (digital). Samples

across time are not
continuous, they are discrete.
Sample times
 Discrete-Time or Digital Signals
To be stored or processed in a digital computer, a
continuous analog signal be converted to a series of
numbers through a process known as sampling.

A continuous
analog signal, x(t), is sliced
up into a sequence of
digital numbers usually at Analog
equal time intervals, Ts. to
Digital

Ts is known as the sampling

interval.
 Digital Signals

• These sequential numbers represent the value of the

analog signal at a discrete points in time determined
by the sample interval, Ts:

t = nTs = n/fs

Where fs is the sampling frequency, fs = 1/Ts and

where n is the position of the number in the sequence,
or the amount of samples.

• Usually this series of numbers would be stored in

sequential memory locations (called and array or
vector) with x[1] followed by x[2], then x[3], and so on.
Sampled or Digitized Signals
• A sampled time signal, also referred to as a digitized
signal or simply digital signal, can be easily stored in a
digital computer.

• A digital signal, x[k], is just a series of numbers:

x[k]= x1, x2, x3, . . . xN.
 Digital or Analog?
Signal computed Resolution is the smallest possible change that a
in MATLAB Online sensor can perceive.
last week.
Time is discrete.

Ts = 0.01 s
fs = 1/Ts; fs = 100 Hz

How can we increase our signal’s time

resolution?

By increasing the sampling frequency (fs).

t (s)
 Amplitude Slicing: Quantization
• Since digital numbers can only represent discrete or
specific amplitudes, the analog signal must also be
sliced up in amplitude.

To digitize or
sample an
analog signal
requires
slicing the
signal in two
ways: in time
and in
amplitude.
Analog-to Digital Conversion
Conversion between the analog and digital domain is done
with special hardware devices know as analog-to-digital
or digital-to-analog converters.

Continuous (analog) signal Discrete (digital) signal

Digital to Analog

Analog-to-digital
Analog-to Digital Conversion

Compares analog value to specific

(discrete) values, and assigns 1 or 0
depending on the outcome.

Information is stored in bits

Analog-to-digital Conversion

Example 1-1 A 12-bit analog-to-digital converter (ADC)

advertises an accuracy of ± the least significant bit (LSB).
If the input range of the ADC is 0 to 10 volts, what is the
accuracy of the ADC in analog volts?

Solution: If the input range is 10 volts then the analog voltage

represented by the LSB would be:

V max 10 10
VLSB = Nu bits = 12 = = .0024 volts
2 2 4096

Hence the accuracy (resolution) would be ± .0024 volts.

Resolution can increase by increasing the number of bits in the ADC.
 Noise and Variability
Noise is what you do not want and signal is what you do want
(noise is unwanted variability). Noise often limits the usefulness
of a signal and the diagnostic accuracy of a medical device.

Sources of Variability

Source Cause Potential Remedy

Physiological Measurement only indirectly Modify overall approach
variability relate to variable of interest
Environmental or Other source of similar energy Noise cancellation and/or
interference form transducer design
Artifact Transducer responds to other Transducer design
energy forms
Electronic Thermal or shot noise Transducer or electronic
design
Sources of Variability (cont)

• Physiological variability is due to the fact that the

information you desire is based on measurements subject
to biological influences other than those of interest. (e.g.
assessment of respiratory function based on the measurement of blood pO2)
• Environmental noise can come from sources external or
internal to the body. (e.g. fetal ECG is corrupted by the mother’s ECG.)
• Measurement artifact is produced when the biotransducer
responds to energy modalities other than those desired. (e.g.
electrodes susceptible to motion artifact.)
• Electronic noise is the only noise that has well know
source and characteristics. (Sources are either thermal activity and/or
semiconductor junctions.)
EEG contamination
• The images on the left shows
in the top panels, a clean
EEG signal and its power
spectrum.

• All the panels below, show

EEG signal contaminated
with different noise
sources, and how they
affect the power spectrum.
EEG contamination

Slow “up” and “down”

movements mimick the
eye motion.

Fast “spikes” reflect the

heartbeat trace.

• Sometimes, the contamination can also be observed in the

time domain.
Electronic Noise
Spectrum of White Noise
0.03 Frequency Spectrum of Electronic Noise

0.025

0.02

0.015

0.01

0.005

0
20 40 60 80 100 120
Relative Frequency

Electronic noise is well defined. It has energy at all frequencies.

This leads to the term white noise.
To give a number to the amount of noise present, a range of
frequencies (i.e., bandwidth) must be specified.
 Time Domain Measurements
Most signal analysis requires quantitative measurement;
determining some numbers that summarize its medically useful
features.

The signal on the right has higher overall values, but because it has less
fluctuation it appears less energetic than the one on the left.
The two most basic signal measurements, the mean and rms value,
quantify these two differences.
Time Domain Measurements: Mean

One of the most straightforward of signal measurements is

the assessment of its average or mean value. For an
analog signal, x(t)
1 T
x(t) =  x(t) dt
T 0
In the digital domain, integration becomes summation
and the continuous signal, x(t), becomes the a discrete
signal xn, and T becomes N:
N
1
xavg = x =
N
x
n=1
n
 Basic Measurements: RMS
Although the average value is a basic property of a signal,
it does not provide any information about the variability
of the signal.

The root mean squared (rms) value is a measurement

that includes both the signal’s variability and its average.
1
1 T 2  2
x(t)rms =   x(t) dt 
T 0 

The rms is calculated by following the acronym

(rms) backward: first squaring the signal (S), then
taking its average (M), and finally taking the square
root of this average (R).
 Basic Measurements: RMS

The discrete form of the equation is obtained by

following the same rules used for the mean:
1
1 N
 2
xrms =   xn 
2

N n=1 

Example 2.1 Find the RMS value of the sinusoidal signal:in both
continuous an discrete domain.

(
x( t ) = A sin(p t) = A sin 2 t
T )
Analytical (i.e., continuous) solution Since this signal is periodic, with
each period the same as the previous one, it is sufficient to apply the
RMS equation over a single period.
Basic Measurements: Variance

• A statistical measure related to the RMS value is the

“variance,” σ2.
• The variance is a measure of signal variability
irrespective of its average.
• The calculation of variance for both discrete and
continuous signals is given as:
T
1 
 =2

T0

2

 x ( t ) - x  dt

N
1
 =
2

N-1 n=1
(xn - x)
2
Dividing by N-1 gives a
better estimate of variance
if the signal is Gaussian.
Basic Measurements: Standard Deviation
The “standard deviation” is another measure of a signal’s
variability and is simply the square root of the variance:

1
1 T   1 N 2
1


2 2
 =    x (t ) - x dt 
2
 =  (xn - x) 
T 0   N-1 n=1 

The average standard deviation of multiple sources is

equal to the average of the individual standard deviation
divided by N.

When multiple noise sources are present, the variances add:

N
1
 2 =  k 2
N k=1

Averaging can be used to enhance SNR by √N

Ensemble Averaging

• Averaging can be applied to entire signals, a

technique known as “ensemble averaging.”

• Ensemble averaging is a simple yet powerful signal

processing technique for reducing noise when
multiple observations of the signal are possible.

• Such multiple observations could come from multiple

sensors, but in many biomedical applications the
multiple observations come from repeated responses
to the same stimulus.
Ensemble Averaging
• In “ensemble averaging,” a group of time responses are
averaged together on a point-by-point basis.

• There are two essential requirements for the application of

ensemble averaging for noise reduction:
1. the ability to obtain multiple observations-
2. a reference signal closely time-linked to the response.

• When the individual signals are aligned in time, an

ensemble average signal is constructed by averaging over
corresponding points from the individual signals.

The next slide shows 3 noisy sine waves and their ensemble
average (the X symbol shows data samples that correspond
time.
 Example of Ensemble Averaging

The
individual
sine waves
have first
been aligned
before
averaging .
The averaged
response is
actually the
ensemble
average of
200 noisy
sinusoids.