Efficient codec for electrical signals

Abstract

The present invention provides a method for compressing a signal, the method comprising: obtaining a main signal via a signal recording module; modeling a model signal of the main signal via a processor by: obtaining a sampled signal via the processor; obtaining a windowed signal via the processor; and extracting, via the processor: a fundamental frequency waveform having a fundamental magnitude and a fundamental phase; and at least one harmonic frequency waveform having a harmonic magnitude and a harmonic phase; wherein the model signal comprises the fundamental frequency waveform and the at least one harmonic frequency waveform; calculating, via the processor, an error signal between a reconstructed signal and the main signal; determining, via the processor, an optimal gain according to at least: an averaging step providing an average, a predefined threshold, and a scaling signal, wherein the scaling signal is a historical error signal scaled by iteratively: averaging a difference between the error signal and the scaling signal, wherein the optimal gain comprises a predefined gain when the average satisfies the predefined threshold; determining, via the processor, an index from a residual signal by: determining the residual signal; vector quantizing the residual signal; and indexing the vector quantized residual signal; synthesizing, via the processor, a compressed signal, wherein the compressed signal comprises: the fundamental phase; the fundamental magnitude; the harmonic phase; the harmonic magnitude the optimal gain; the index. Thus, the compression method preferably overcomes problems associated with current compression techniques, and provides a suitable technique for compressing a signal that can be used to infer a type of load on a circuit.

Description

Technical Field

[0001] This invention relates to methods for compressing and decompressing electrical signals. Background Technology

[0002] Energy management systems capture crucial data related to voltage and current waveforms for end-use energy or load monitoring. The data captured by an energy management system requires a high sampling rate to infer the electrical load (i.e., the type of machine or device present in the circuit) or predict when maintenance of equipment in the circuit is needed.

[0003] In certain scenarios, it may be necessary to transmit this data over a considerable distance via a network. Typically, these networks are constrained, making the data rate required to transmit voltage and current waveforms potentially too high. Even for state-of-the-art data transmission methods such as fiber optics, the required data rate is often too large.

[0004] Therefore, a technique to reduce data size is needed.

[0005] Current methods for signal compression are based on minimizing the average error between the initial and reconstructed signals. However, inferring the type of machine or device is based on higher-order components of the signal. Therefore, current compression methods that minimize, for example, the mean square error may remove these higher-order components that allow the system to extract information about the electrical load.

[0006] There are other signal compression methods where the signal is compressed based on a psychoacoustic model. However, these methods are not suitable for inferring the electrical load from the electrical signal.

[0007] Therefore, an improved method for compressing and decompressing signals is needed. Summary of the Invention

[0008] According to a first aspect of the present invention, a method for compressing a signal is provided, the method comprising: acquiring a main signal via a signal recording module; modeling a model signal of the main signal via a processor by: acquiring a sampled signal via the processor; acquiring a windowed signal via the processor; extracting via the processor: a fundamental frequency waveform having a fundamental value, a fundamental phase, and a fundamental frequency; and at least one harmonic frequency waveform having harmonic values, harmonic phases, and harmonic frequencies; wherein the model signal includes the fundamental frequency waveform and the at least one harmonic frequency waveform; calculating an error signal between a reconstructed signal and the main signal via the processor; and determining an optimal gain via the processor according to at least the following: The process includes an averaging step, a predefined threshold, and a scaling signal, wherein the scaling signal is a historical error signal scaled by iteratively performing the following steps to a predefined gain: averaging the difference between the error signal and the scaling signal, wherein the optimal gain includes the predefined gain when the average value satisfies the predefined threshold; determining an index from the residual signal via the processor by: determining the residual signal; vector quantizing the residual signal; and indexing the vector-quantized residual signal; and synthesizing the compressed signal via the processor, wherein the compressed signal includes: the fundamental phase; the fundamental magnitude; the fundamental frequency; the harmonic phase; the harmonic magnitude; the harmonic frequency; the optimal gain; and the index. In this way, the main signal can be compressed to produce a compressed signal, which includes components that allow the signal to be transmitted at a lower data rate than the main signal. Those skilled in the art should note that each harmonic frequency waveform in at least one harmonic frequency waveform includes a harmonic phase, a harmonic magnitude, and a harmonic frequency.

[0009] The term "model signal" will be understood by those skilled in the art to refer to a signal that has been modeled. The term "fundamental quantity" will be understood by those skilled in the art to refer to the quantity of the fundamental signal, the term "fundamental phase" will be understood to refer to the phase of the fundamental signal, and the term "fundamental frequency" will be understood to refer to the frequency of the fundamental signal. Similarly, the terms "harmonic quantity," "harmonic phase," and "harmonic frequency" will be understood to refer to the quantity, phase, and frequency of the harmonic signal, respectively.

[0010] Preferably, the sampled signal is acquired by sampling the main signal at a Nyquist rate via a processor. In this way, the main signal can be compressed by the processor by sampling the main signal at the lowest sampling rate without introducing errors such as aliasing into the compressed signal.

[0011] Those skilled in the art will understand that other sampling rates are conceivable, where the sampling rate is any rate suitable for capturing the master signal.

[0012] Preferably, the windowed signal is obtained by applying a window function to the sampled signal via a processor. More preferably, the window function is a Hamming window. In this way, spectral leakage can be reduced.

[0013] Those skilled in the art will understand that any window function suitable for minimizing the effects of spectral leakage can be used.

[0014] Preferably, the fundamental frequency and at least one harmonic frequency are extracted by applying a Fast Fourier Transform to the windowed signal via a processor. More preferably, the fundamental frequency and six harmonic frequencies are extracted. In this way, the main signal can be further compressed into seven frequency components while still retaining more information than if only the fundamental frequency waveform were extracted.

[0015] Those skilled in the art will understand that any number of harmonic frequencies can be extracted. In this way, a higher number of harmonic frequencies can capture the main signal more accurately, while a lower number of harmonic frequencies can further compress the main signal.

[0016] Preferably, the reconstructed signal is based on the model signal. In this way, the reconstructed signal can represent a compressed version of the main signal, including the magnitude, phase, and frequency of the fundamental signal, as well as the magnitude and phase of each harmonic signal in the harmonic signals.

[0017] Preferably, the error signal is calculated by subtracting the reconstructed signal from the main signal. In this way, the difference between the reconstructed signal and the main signal can be obtained, which indicates the loss of signal information.

[0018] Preferably, determining the optimal gain includes the following iterative steps: i. encoding the scaled signal via a processor; wherein the scaled signal is encoded by multiplying a historical error signal by a predefined gain via a processor; ii. subtracting the scaled signal from the error signal via a processor to produce a difference; iii. calculating the average value of the resulting signal via a processor; iv. comparing the average value with a predefined threshold via a processor; and v. repeating steps i. to iv., wherein the predefined gain is the new gain. In this way, the delay component of the error signal can be extracted, which represents the cyclic signal common to the error signal across multiple time periods. Further advantageously, an optimal gain value can be achieved, thereby providing the optimal compression ratio.

[0019] In a preferred embodiment, the historical error signal is an error signal from a previous time interval stored in a data repository. In this way, error signals from previous time periods can be stored to extract the delay component from the error signal.

[0020] Preferably, the new gain is calculated by the processor using a stochastic descent algorithm. In this way, the optimal gain value can be achieved in fewer iterations in steps i to v.

[0021] Those skilled in the art will understand that any algorithm suitable for convergence to the optimal gain can be used.

[0022] Preferably, the residual signal is encoded by determining the difference between the error signal and the historical error signal. In this way, a signal that includes information related to the difference between the error signals at two different time intervals can be created.

[0023] Preferably, the residual signal is vector-quantized relative to a predefined codebook, which is stored in a data repository. Advantageously, vector quantization of the residual signal compresses it. More preferably, the predefined codebook includes an index configured to provide an optimal compression ratio. Advantageously, the vector quantization points (i.e., the indexes) can be selected such that minimal data loss occurs during the vector quantization step. In another preferred embodiment, the predefined codebook is trained as more main signals are compressed. In this way, the predefined codebook can improve the indexes as more main signals are compressed.

[0024] Preferably, the residual signal is incrementally encoded to generate incrementally encoded coefficients. In this way, the residual signal is stored as the difference between the error signal and the historical error signal. More preferably, if the predefined codebook has not yet been trained to a suitable level, the residual signal is only incrementally encoded.

[0025] Preferably, the compressed signal is stored in a data repository or transmitted via a transmitter. In this way, the compressed signal can be stored or transmitted.

[0026] According to a second aspect of the present invention, a method for decompressing a compressed signal is provided, the method comprising: extracting a residual signal via a processor; and reconstructing a main signal via the processor.

[0027] In some implementations, the compressed signal includes: fundamental phase; fundamental magnitude; fundamental frequency; harmonic phase; harmonic magnitude; harmonic frequency; optimal gain; and an index. Advantageously, these components of the compressed signal can be used to reconstruct the compressed signal.

[0028] In another implementation, the compressed signal includes: fundamental phase; fundamental magnitude; fundamental frequency; harmonic phase; harmonic magnitude; harmonic frequency; optimal gain; and incremental coding coefficients. Advantageously, these components of the compressed signal can be used to reconstruct the already compressed signal.

[0029] In some implementations, extracting the residual signal includes: comparing an index with a predefined codebook stored in a data repository via a processor; and reconstructing the residual signal based on the index via the processor. In this way, the index can be found in the predefined codebook and subsequently used to reconstruct the residual signal.

[0030] In another implementation, extracting the residual signal includes: decoding the incremental coding coefficients via a processor; and reconstructing the residual signal via the processor based on the decoded incremental coding coefficients. In this way, the incremental coding coefficients can be used instead of an index to reconstruct the residual signal.

[0031] Preferably, reconstructing the main signal includes: adding the residual signal to a historical residual signal multiplied by the optimal gain via a processor to generate an added residual signal; reconstructing the sinusoidal component of the main signal via a processor based on the following: fundamental phase; fundamental magnitude; fundamental frequency; harmonic phase; harmonic magnitude; and harmonic frequency; and adding the added residual signal to the sinusoidal component via a processor to reconstruct the main signal. In this way, the main signal can be reconstructed using components of a compressed signal.

[0032] Preferably, the historical residual signal is a residual signal from a previous time interval stored in a data repository.

[0033] According to a third aspect of the present invention, a method for compressing and decompressing signals is provided, comprising the methods according to the first aspect and the second aspect of the present invention.

[0034] According to a fourth aspect of the present invention, a system for compressing a signal is provided, the system comprising: a signal recording module configured to acquire a master signal; a data storage repository including: a historical error signal; a predefined codebook; and a compressed signal; a compression unit configured to compress the master signal, the compression unit including a processor configured to: model a model signal; calculate an error signal; determine an optimal gain; determine an index; determine incremental coding coefficients and synthesize the compressed signal; and a transmitter configured to transmit the compressed signal.

[0035] Preferably, the processor is further configured to: acquire a sampled signal; acquire a windowed signal; apply a fast Fourier transform to the windowed signal; extract a fundamental frequency waveform and at least one harmonic frequency waveform; calculate a reconstructed signal; subtract the reconstructed signal from the main signal; multiply the historical error signal by the predefined gain to obtain a scaled signal; average the difference between the error signal and the scaled signal; subtract the scaled signal from the error signal to generate a resulting signal; calculate the average value of the resulting signal; compare the average value with the predefined threshold; repeat the step of extracting the delay component using a new gain until the average value meets the predefined threshold; calculate the new gain using a random descent algorithm; perform vector quantization on the residual signal relative to the predefined codebook; perform incremental encoding on the coefficients; and synthesize the compressed signal by combining the following: the fundamental phase; the fundamental magnitude; the fundamental frequency; the harmonic phase; the harmonic magnitude; the harmonic frequency; the optimal gain; the index or the incremental encoding coefficient.

[0036] According to a fifth aspect of the present invention, a system for decompressing a compressed signal is provided, the system comprising: a receiver configured to receive a compressed signal including: a fundamental phase; a fundamental magnitude; a fundamental frequency; a harmonic phase; a harmonic magnitude; a harmonic frequency; an optimal gain; and indexed or incremental coding coefficients; a data repository including: a predefined codebook; and a historical residual signal; and a decompression unit including a processor configured to: extract the residual signal; multiply the historical residual signal by the optimal gain; add the residual signal to the multiplied historical residual signal to generate an added residual signal; reconstruct the main signal; and add the main signal to the added residual signal. Detailed Implementation

[0037] Specific implementation schemes will be described by way of example only, with reference to the accompanying drawings, wherein:

[0038] Figure 1 A schematic diagram of the energy management system in the circuit is shown;

[0039] Figure 2 A schematic diagram of a compression method for compressing a signal according to a first aspect of the present invention is shown;

[0040] Figure 3 A schematic diagram of the compression unit is shown; and

[0041] Figure 4 A schematic diagram of a decompression method for decompressing a compressed signal according to a second aspect of the present invention is shown.

[0042] See Figure 1The diagram shows a schematic of the energy management system 100 in circuit 101. The energy management system 100 includes a signal recording module 102, a data storage unit 104, a compression unit 106, a transmitter 108, and a processor 110. The data storage unit 104 includes historical error signals 105, predefined gains 107, predefined thresholds 109, and a codebook 111. The codebook 111 includes multiple indices.

[0043] In use, and for reference Figure 2 This illustrates a compression method 200 for compressing signals according to a first aspect of the present invention.

[0044] At step 202, the signal recording module 102 records the main signal 203. The main signal can be a current or voltage signal corresponding to the circuit 101.

[0045] At step 204, processor 110 samples the main signal at the Nyquist rate to generate sampled signal 205.

[0046] In step 206, the processor 110 applies a Hamming window to the sampled signal 205, thereby generating a windowed signal 207.

[0047] At step 208, the processor 110 applies a fast Fourier transform to the windowed signal 207 to generate the spectrum 209 of the windowed signal 207.

[0048] In step 210, processor 110 extracts the fundamental signal and the first six harmonic signals from spectrum 209.

[0049] At step 212, processor 110 extracts the fundamental frequency, fundamental phase, and fundamental frequency from the fundamental signal of spectrum 209. Processor 110 further extracts the harmonic frequency, harmonic phase, and harmonic frequency from each of the six harmonic signals in spectrum 209. The fundamental frequency, fundamental phase, harmonic frequency, and harmonic phase are stored in data repository 104.

[0050] In step 214, the processor 110 uses the fundamental wave value, fundamental wave phase, fundamental wave frequency, harmonic value, harmonic phase, and harmonic frequency to reconstruct the model signal 215.

[0051] At step 216, processor 110 subtracts model signal 215 from main signal 205 to generate error signal 217.

[0052] At step 218, processor 110 initiates an iterative process by multiplying the historical error signal 105 by a predefined gain 107. Multiplying by the predefined gain 107 generates a scaled signal 219.

[0053] At step 220, processor 110 subtracts scaling signal 219 from error signal 217. Subtracting scaling signal 219 from error signal 217 produces the resulting signal 221.

[0054] At step 222, processor 110 calculates the average of the differences and compares the average with a predefined threshold 109.

[0055] At step 224, if the average value meets the predefined threshold 109, then step 226 does not occur, and the predefined gain is output as the optimal gain 225. If the average value does not meet the predefined threshold 109, then step 226 occurs.

[0056] At step 226, the predefined gain is adjusted to the new gain 227 using a random descent algorithm, and steps 218 to 224 are repeated with the predefined gain 107 being the new gain 227.

[0057] At step 228, processor 110 subtracts error signal 217 from historical error signal 105 to generate residual signal 229.

[0058] At step 230, processor 110 compares residual signal 229 with indices in codebook 111 to vector quantize residual signal 229 into a plurality of residual indices 231.

[0059] At step 232, processor 110 synthesizes a compressed signal 233, which includes the fundamental frequency, fundamental phase, fundamental frequency, harmonic frequencies, harmonic phases, harmonic frequencies, optimal gain 225, and residual index 231. It should be noted that the compressed signal 233 includes the harmonic frequency, harmonic phase, and harmonic frequency of each of the six harmonic waveforms.

[0060] At step 234, transmitter 108 sends compressed signal 233.

[0061] refer to Figure 3 The diagram shows a decompression unit 300.

[0062] The decompression unit 300 includes a processor 302, a receiver 304, and a data storage unit 306. The data storage unit 306 includes a codebook 308 and a historical residual signal 312. The codebook 308 is substantially similar to... Figure 1 The codebook 111 is described above, and includes multiple indexes.

[0063] In use, and for reference Figure 4 The diagram shows a decompression method 400 for decompressing a compressed signal 233 according to a second aspect of the present invention.

[0064] At step 402, receiver 304 receives compressed signal 233 and stores compressed signal 233 in data storage 306. Compressed signal 233 includes fundamental frequency, fundamental phase, fundamental frequency, harmonic frequency, harmonic phase, harmonic frequency, optimal gain 225, and residual index 231.

[0065] At step 404, processor 302 compares residual index 231 with an index in codebook 308 and generates reconstructed residual signal 405 based on the index that matches residual index 231. Reconstructed residual signal 405 will be similar to residual signal 229, but with some information lost.

[0066] At step 406, the processor multiplies the historical residual signal 312 by the optimal gain 225 to generate the scaled historical residual signal 407.

[0067] At step 408, the processor adds the reconstructed residual signal 405 to the scaled historical residual signal 407 to generate the summed residual signal 409.

[0068] In step 410, the processor combines the fundamental wave value, fundamental wave phase, and fundamental wave frequency to generate a fundamental wave sinusoidal signal 411.

[0069] At step 412, the processor combines the harmonic magnitude, harmonic phase, and harmonic frequency to generate a harmonic sinusoidal signal 413.

[0070] In step 414, the processor adds the fundamental sine signal 411 to the harmonic sine signal 413 to generate a sine signal 415.

[0071] At step 416, the processor adds the sinusoidal signal 415 to the summed residual signal 409 to generate the decompression signal 417. The decompression signal 417 is essentially similar to the main signal 203, but some information is lost.

[0072] It should be understood that the above embodiments are given by way of example only, and various modifications can be made to the embodiments without departing from the scope of the invention as defined by the appended claims. For example, the main signal can be a current, voltage, or any other circuit indicator. Furthermore, any number of harmonic signals can be extracted from the main signal. Furthermore, the main signal can be sampled at any sampling rate, and any window function can be applied to the sampled signal. Furthermore, any suitable algorithm can be used to converge to the optimal gain value.

Claims

1. A method for compressing a main signal, the method comprising: The main signal is acquired via a signal recording module; The processor models the model signal of the main signal using the following methods: The processor acquires a sampled signal from the main signal; The processor acquires a windowing signal from the sampled signal; as well as The processor extracts from the windowed signal: A fundamental frequency waveform that includes the fundamental wave value, fundamental wave phase, and fundamental wave frequency; A waveform with at least one harmonic frequency, having harmonic magnitude, harmonic phase, and harmonic frequency; and The model signal is reconstructed based on the fundamental frequency waveform and the at least one harmonic frequency waveform; The processor calculates the error signal between the model signal and the main signal. The processor determines the optimal gain based on at least the following: an averaging step providing an average value, a predefined threshold, and a scaling signal, wherein the scaling signal is a historical error signal predefined by iteratively performing the following: The difference between the error signal and the scaling signal is averaged, wherein when the average value satisfies the predefined threshold, the optimal gain includes the predefined gain; The residual signal is generated by subtracting the error signal from the historical error signal. The residual signal vector is quantized into multiple residual indices; as well as The processor synthesizes a compressed signal, wherein the compressed signal includes: The fundamental phase; The fundamental frequency value; The fundamental frequency; The harmonic phase; The harmonic values; The harmonic frequency; The optimal gain; and The residual index.

2. The method according to claim 1, wherein: The sampled signal is obtained by sampling the main signal at the Nyquist rate via the processor; The windowed signal is obtained by applying a window function to the sampled signal via the processor; and The fundamental frequency waveform and the at least one harmonic frequency waveform are extracted by applying a fast Fourier transform to the windowed signal via the processor.

3. The method of claim 1, wherein determining the optimal gain comprises the following iterative steps: i. The scaling signal is encoded via the processor; The scaling signal is encoded by multiplying the historical error signal by the predefined gain via the processor; ii. The difference is generated by subtracting the scaling signal from the error signal via the processor; iii. The average value of the difference is calculated via the processor; iv. Compare the average value with the predefined threshold via the processor; and v. Repeat steps i. to iv., where the predefined gain is the new gain.

4. The method of claim 3, wherein the new gain is adjusted via the processor using a random descent algorithm.

5. The method of claim 1, wherein the residual signal is vector-quantized relative to a predefined codebook, wherein the predefined codebook is stored in a data repository of the energy management system.

6. The method of claim 1, wherein the residual signal is incrementally encoded to generate incrementally encoded coefficients.

7. A method for decompressing a compressed master signal obtained by the method of claim 1, the method comprising: The processor extracts the fundamental phase, fundamental magnitude, fundamental frequency, harmonic phase, harmonic magnitude, harmonic frequency, optimal gain, and residual index. as well as The processor reconstructs and decompresses the main signal based on the fundamental phase, the fundamental magnitude, the fundamental frequency, the harmonic phase, the harmonic magnitude, the harmonic frequency, the optimal gain, and the residual index.

8. The method of claim 7, further comprising reconstructing the residual signal by: The processor compares the residual index with a predefined codebook, wherein the predefined codebook is stored in a data repository; and The processor reconstructs the residual signal based on the residual index relative to the predefined codebook.

9. The method according to claim 7, wherein reconstructing the decompression master signal comprises: The processor adds the residual signal to the historical residual signal multiplied by the optimal gain to generate a summed residual signal. The processor reconstructs the sinusoidal component of the decompressed main signal based on the following: Fundamental phase; Fundamental frequency value; Fundamental frequency; Harmonic phase; Harmonic values; and Harmonic frequencies; The processor adds the summed residual signal to the sinusoidal component to reconstruct the decompressed main signal.

10. The method of claim 9, wherein the historical residual signal is a previous residual signal from a previous time interval stored in a data repository.

11. A system for compressing a main signal, the system comprising: A signal recording module, configured to acquire the main signal; Data repository, the data repository comprising: Historical error signals; Predefined codebook; and A processor configured to perform the compression method according to claim 1; and A transmitter configured to send a compressed master signal.

12. A system for decompressing a compressed master signal, the system comprising: A receiver configured to receive the compressed master signal; Data repository, the data repository comprising: Predefined codebook; and Historical residual signals; and A decompression unit, the decompression unit including a processor configured to perform the decompression method according to claim 7.

Images