Fundamentals of Signals, DSP and Applica7ons in m- Health. By Deepta Rajan FSE Oct 10, 2013.

Size: px

Start display at page:

Download "Fundamentals of Signals, DSP and Applica7ons in m- Health. By Deepta Rajan FSE Oct 10, 2013."

Augustine McLaughlin
6 years ago
Views:

1 Fundamentals of Signals, DSP and Applica7ons in m- Health By Deepta Rajan FSE Oct 10, 2013.

2 Outline Signals What are they? Fourier Transform - T/F domain Challenges in Signal Processing The AJDSP App Types of Signals Sensor Signals audio, accelerometer, video Physiological Signals ECG, PPG, GSR Mobile Health Monitoring M- Health apps using AJDSP

3 Signals What are they? Any informa7on is a signal. In the real- word, a signal is any quan7ty that provides informa7on and varies over 7me or space. Examples: speech signal, images, video, weather in Tempe, engine noise, ECG, voltage, current, radio waves, music, moving cars etc! Image source - hyp://en.wikipedia.org/wiki/file:signal_processing_system.png

4 Defini7ons Time period - Time dura7on of one cycle or the number of seconds per cycle. Ex: Time between successive heart beats. Frequency Number of cycles per second. Ex: Number of heart beats per second. Amplitude Magnitude of the pulse. Ex: In a sound wave, it s the magnitude of air pressure. Phase Change in ini7al sinusoidal angle. Sine and Cosine func7ons in- phase or out- of- phase? Bandwidth In a set of con7nuous frequencies, it is the difference between the upper and lower frequency.

5 Sinusoidal Signals sin(! 0 t)=sin(2 f 0 t)=sin 2 t T 0! 0 =2 f 0 =2 t, f 0 = 1 T 0 T 0 Frequency Time Period T 0 0 t EEE Spring 2013 Mahesh K. Banavar II- 5

6 Sinusoidal Signals (2) cos(! 0 t) = cos(2 f 0 t) = cos 2 t T 0! 0 =2 f 0 =2 t T 0, f 0 = 1 T 0 T 0 0 t EEE Spring 2013 Mahesh K. Banavar II- 6

7 Frequencies x 1 (t) =sin(! 1 t) T 1 x 2 (t) =sin(! 2 t) T 2 x 3 (t) =sin(! 3 t) T 3! 1 <! 2 <! 3 T 1 >T 2 >T 3 EEE Spring 2013 Mahesh K. Banavar II- 7

8 Unit Step Func7on Con7nuous- 7me Unit Step Func7on u(t) = ( 1 t 0, 0 else u(t) 1... Defined along a con7nuum of 7me. It s a set of real numbers. Also referred to as Analog Signals. 0 t EEE Spring 2013 Mahesh K. Banavar II- 8

9 Unit Step Func7on Discrete- 7me Unit Step Func7on u(n) = ( 1 n 0, 0 else 1 u(n) 0... n It s a set of integers. Also referred to as Digital Signals. EEE Spring 2013 Mahesh K. Banavar II- 9

(a) (b) (b) Quan7za7on Mapping the points on the con7nuous signal to

10 Sampling and Quan7za7on (a) Sampling Conversion from con7nuous- 7me to discrete- 7me by measuring signal values every few seconds. (a) (b) (b) Quan7za7on Mapping the points on the con7nuous signal to some values. A way of rounding off. Ex: Going from Analog to Digital signal.

11 Time domain and Frequency Domain Time Domain Frequency Domain Amplitude vs Time Amplitude vs Frequency Observa7on: Slowly 7me- varying signals have low- frequency content. Signals with abrupt changes in amplitude have high frequency components. The frequency content of a signal can be es7mated using Fourier methods. EEE Spring 2013 Mahesh K. Banavar I- 11

12 Example Speech signals 1.0 Time domain speech segment TAPE TIME: fundamental frequency Formant Structure Amplitude 0.0 Magnitude (db) Time (ms) Frequency (KHz) 1.0 Time domain speech segment 40 TAPE TIME: Amplitude 0.0 Magnitude (db) Time (ms) Frequency (KHz) EEE Spring 2013 Mahesh K. Banavar I- 12

13 Prototype Filters Low pass filter High Pass filter H LP (!) H HP (!)! c! c!! c! c! Band- pass filter Band- stop filter H BP (!) H BS (!)! c2! c1! c1! c2! c2! c1! c1! c2!! EEE Spring 2013 Mahesh K. Banavar XIV- 13

14 Signal Processing Challenges Denoising Source separa7on Speech recogni7on Peak detec7on Demodula7on Face recogni7on Auto tuning And many more! J

Android- Java- DSP (AJDSP) How to use? Watch quick demo hyp://www.youtube.

15 Android- Java- DSP (AJDSP) How to use? Watch quick demo hyp:// v=e2xpehprzs0&list=plde6647e29748aade Signal processing modules designed as blocks. Connect blocks to set up simula7ons in the workspace. Topics: filter design Convolu7on mul7rate signal processing the FFT discrete wavelet transform.

16 Sensor Signals Audio Signal Video Signal Images are 2D signals Accelerometer Signal

17 Mobile Health Monitoring Using mobile devices to es7mate physiological parameters by monitoring the corresponding physiological signals. Several apps available in the market. Examples: Heart Rate Blood Pressure Calorie Counter Pedometer Respiratory Rate Oxygen Satura7on in the blood

18 m- Health Applica7ons Speech Denoising Heart Rate Es7ma7on Oxygen Satura7on (SpO2) Step Counter

19 Speech Denoising

20 ECG Signals Image source: F. T. Sun, C. Kuo, H. T. Cheng, S. Buthpi7ya, P. Collins and M. Griss, Ac7vity- Aware Mental Stress Detec7on Using Physiological Sensors, In Mobile Compu7ng, Applica7ons, and Services, pp , Springer Berlin Heidelberg, 2012.

21 ECG Signals

22 Photoplethysmogram (PPG) Principle: the absorp7on of light by a medium causes the incident light intensity to drop exponen7ally, and the amount of absorp7on is related to the wavelength of light. The PPG corresponds to the change in blood volume with 7me. The PPG typically comprises of two components, a pulsa7le component and a constant component obtained from the measurement of changes in light absorp7on by the skin.

23 HbO2 and Hb Absorp7on Spectra

24 Heart Rate Parameter Es7ma7on Oxygen Satura7on (SpO2)

25 Accelerometer

27 References filter which discards high- frequency events while retaining the general shape of the signal was desired. The energy filter first considers the total energy inside the window. To find this quan7ty, the energy of each of the bins (except the DC) is summed. Again, by Parseval s theorem, this sum is equivalent to the total energy of the window less its mean. The total energy is mul7plied by a parameter to the filter, the rela7ve energy threshold, to ayain a threshold energy Why are we not elimina7ng the DC value?

28 SHIMMER Applica>ons Kinema7c - Accelerometers, gyroscopes and magnetometers are included. Physiological - Galvanic Skin Response (GSR), Electrocardiogram (ECG), and Electromyography (EMG) sensors. Ambient sensing - Comprises of temperature and light sensors. AJDSP interfaces with the SHIMMER Accelerometer, ECG, and GSR through Bluetooth. Examples: Energy expenditure es7ma7on in Rheumatoid Arthri7s pa7ents, ECG compression, stress detec7on, mo7on analysis of pa7ents with Parkinson s disease, stroke and epilepsy.

29 Publica>ons Rajan, D. Ranganath, S, Banavar, M. Spanias,A. Health Monitoring Laboratories by interfacing Physiological Sensors to Mobile Android Devices IEEE Fron(ers in Educa(on Conference, Rajan, D. Kalyanasundaram, G. Hu, S. Banavar, M. Spanias,A. Development of mobile sensing apps for DSP applica7ons IEEE DSP Workshop, Ranganath, S. Thiagarajan, J.J.,Ramamurthy, K.N., Hu,S. Banavar, M. Spanias, A. Work in progress: Performing signal analysis laboratories using Android devices. IEEE Fron(ers in Educa(on Conference, Ranganath, S. Thiagarajan, J.J.,Ramamurthy, K.N., Hu,S. Banavar, M. Spanias, A. Undergraduate Signal Processing Laboratories for the Android Opera7ng System. ASEE, Ranganath, S. Rajan, D. Thiagarajan J.J.,Ramamurthy, K.N., Hu,S. Banavar, M. Spanias,A Undergraduate Signal Simula7ons and Anima7ons for the Android Opera7ng System. Journal Ar7cle (in prepara7on).

Sensor, Signal and Information Processing (SenSIP) Center and NSF Industry Consortium (I/UCRC)

Sensor, Signal and Information Processing (SenSIP) Center and NSF Industry Consortium (I/UCRC) School of Electrical, Computer and Energy Engineering Ira A. Fulton Schools of Engineering AJDSP interfaces