1. IEEE 21451-001: Extracting Knowledge from sensor signals, first steps
Sponsoring Society and Committee: IEEE Industrial Electronics Society/Industrial Electronics Society Standards Committee (IES/IES)
Sponsor Chair: Victor Huang
Joint Sponsor: IEEE Instrumentation and Measurement Society/TC9 - Sensor Technology (IM/ST) Chair Kang Lee- NIST-USA
Working Group Chair – IES: Gustavo Monte (Universidad Tecnologica Nacional, Argentina)
Working Group Co-Chair – IMS: Ruqiang Yan (SouthEast Univ, China)
Gustavo Monte
Chairman IEEE
Universidad Tecnologica Nacional
Regional Del Neuquen. ARGENTINA

2. IEEE 21451-001: Extracting Knowledge from sensor signals, first steps
Outline
Introduction
Information source for algorithms
Standard structure
Applications
Conclusions
Process
IEEE Recommended Practice for Signal Treatment Applied to Smart Transducers

3. Introduction: What, How and Where?
The four elements that define a network.
Depending on the order of execution and the physical location, totally different interconnection networks are generated.
Process
IEEE Recommended Practice for Signal Treatment Applied to Smart Transducers

4. Introduction: What, How and Where?
The IEEE
Focuses on, or propose solutions to:
How to Sense?
Where to Process?
What to Transmit?
Process
IEEE Recommended Practice for Signal Treatment Applied to Smart Transducers

5. Introduction: The future of smart sensors and Why we need a standard for sensor signals
SCOPE: Sensor networks and IoT
DATA
Smart Sensor
INFORMATION
BANDWIDTH
KNOWLEDGE
Data => Samples
Information=> Preprocessed data without a particular objective.
Knowledge-=> Feature extraction of particular interest.
The lowest data rate is obtained by transmitting questions that require signal processing that have a yes / no response.
Not easy to share
Standards are mandatory

6. Sampling, the key for smart sensing
Uniform sampling only give me the value
IEEE
IEEE

7. Sampling, the key for smart sensing
The sensor signal is described by the union of known segments
Information about the shape is needed from the sampling process
Compare signal trajectory against linear
IEEE

8. Sampling, the key for smart sensing
Give us the necessary information to infer global behavior like an observer does.
IEEE

9. Information structure (MCT)
Three vectors describe the signal: (M,C,T)
M (Mark)={ 1.8,1.7,-0.3…..
C (Class)={g,g,e,d,d,g,e….
T (TempPos)={6,8,12,14,………
Executed in real time
There is no aliasing since the error subspaces are controlled by the interpolation error
IEEE

10. Standard Structure Mark vector Class vector Temp vector
Circular Buffer Timestamp
Samples
Segmentation
Algorithm RTSAL
Class vector
Temp vector
FIRST LAYER
SECOND LAYER
Exponential detection
Signal prediction
Noise detection
MCT
Smart filtering
and compression
Impulsive noise
M’C’T’
MCT
Sinusoidal pattern
MCT
Complex pattern
detection and learning
Tendency estimation
MCT
Mean estimation
User defined application
code
IEEE

11. First Layer Exponential detection Noise detection Impulsive noise MCT
ServiceName
Definition
(Interface description Language -IDL)
None (off)
None
Initialization Command
IDL: UInt16 initialization (in UInt32 samplingPeriod, in UInt8 err, in UInt16 channelID, in UInt32 bufferSize, in UInt16 maxSegLength)
Configuration Command
IDL: UInt16 configuration(in _Boolean exponentialEnabling, in _Boolean noiseEnabling, in _Boolean impulseEnabling, in _Boolean sinusoidalEnabling, in _Boolean tendencyEnabling, in _Boolean meanReliableEnabling, in UInt8 exponentialAmount, in UInt8 noisyPeriod, in UInt8 amountNoisy, in UInt8 impulseThreshold, in UInt16 tendencyWindow, in UInt8 windowMeanEstimation)
RawData
IDL: UInt16 readData(in UInt16 n, out UInt16Array data)
MCTSegments
IDL: UInt16 readMCT(in UInt16 n, out UInt16Array m, out UInt8Array c, out UInt32Array t)
ExponentialDetection
IDL: UInt16 exponentialDetection(in UInt8 exponentialAmount, in UInt16 samplingPeriod, out UInt8 exponentialType, out UInt16 timeSinceExponential, out _Boolean exponentialStable)
FIRST LAYER
Exponential detection
Noise detection
Impulsive noise
MCT
Sinusoidal pattern
Tendency estimation
Mean estimation
IEEE

12. Information structure
void MCT_conversion(void) {
float diff = 0;
char buffer[20];
int puini, last,pu=0;
pu=pu_MCT%NN;
diff=((buffer_samples[pu]+buffer_samples[(pu+k_MCT)%NN])/2-buffer_samples[(pu+k_MCT/2)%NN]);
if(diff==0)
Equal=Equal+1;
last=0; }
else if(diff>0)
Up=Up+1; // could be segment F or D
last=1;
else if(diff<0)
IEEE

13. Applications Is this signal periodic? Iteration=8, N=14
Signal reconstruction using linear interpolation
IEEE

14. Applications The values are normal but not the pattern (Class F )
IEEE

15. Applications
Sinusoid detection with noise. FS=22050, f=345 Hz SNR=7dB. Iterations=10. Interpolation error: 1%. Red: original signal. Green: signal reconstructed from MCT vectors. A new set M’C’T’ is obtained.
IEEE

16. Applications Filtered of a real ECG signal ,Error=2 lsb, iterations=10
IEEE

17. Applications Classification of defective bearing vibration signals
Algorithm based on MCT:
Envelope detection using local maximum and variance of envelope
Classification of defective bearing vibration signals
IEEE

18. Conclusions This standard provides a platform for knowledge sharing.
It propose a new sampling technique that removes redundancy while keep the information structure.
KNOWLEDGE
INFORMATION
DATA
Technological convergence is requiring standards in unexpected fields
IEEE

19. IEEE
Thank you!!