Adaptive filter based on 1-bit digital signal calculation and its application

Abstract

An adaptive filter based on a 1-bit digital signal for operation and its application, the adaptive filter includes a digital signal processing operation module, the digital signal processing operation module uses a 1-bit digital stream signal to directly perform an adaptive filtering operation and then output. The present invention also provides the application of an adaptive filter based on a 1-bit digital signal for operation in adaptive system identification, in F‑x LMS adaptive signal cancellation, and in F‑e LMS adaptive signal cancellation. The present invention can directly perform adaptive filtering operations on a PDM signal without converting to a multi-bit digital amplitude coded signal, can eliminate additional PDM to PCM conversion hardware, reduce complexity, and reduce processing delays. At the same time, all multipliers can be eliminated, greatly reducing the complexity of the entire system and increasing the timing margin of the system clock speed.

Description

Technical Field

[0001] The invention belongs to the field of integrated circuits and relates to an adaptive filter for performing operations based on a 1-bit digital signal and an application thereof. Background Art

[0002] With the development of electronic information technology and integrated circuit technology, more and more fields need to convert analog signals into digital signals for further processing using digital signal processing technology. Therefore, the research and design of digital-to-analog converters and analog-to-digital converters is an important technical category in the field of digital technology, and how to accurately and efficiently convert from the digital domain to the analog domain has always been the most popular technical direction in this field. The more commonly used digital signal representation methods include 1-bit digital stream level signals (such as PDM and PWM) and multi-bit pulse code modulation signals (PCM).

[0003] With the rapid development and popularization of digital silicon-based micro-electromechanical system (MEMS) sensors, the development of sensor technology and corresponding digital signal processing technology has been greatly promoted, making it possible to realize more powerful signal acquisition, conditioning and processing modules with smaller size, lower power consumption and better price.

[0004] For example, the rapidly developing digital silicon microphone technology in recent years can make an analog microphone with integrated analog front-end and digital sampling functions as small as a grain of rice, so that an ordinary smart phone can use several such microphones to form a sensor array, and use signal processing technology to complete advanced sensor array algorithm processing functions such as voice noise reduction and beamforming in real time, which greatly improves the performance and technological level of mobile phones and brings unlimited possibilities for more applications based on silicon micro sensors. In order to further reduce the number of signal lines from silicon micro sensors to the main computing unit, the current popular method is to use I2S bus (3-wire bus) or PDM (2-wire bus). PDM itself is a modulation method between analog and digital signals, which belongs to a type of 1-bit digital stream level signal. By integrating a simple sigma-delta modulator in the sensor, a 1-bit digital signal can be directly output to the processing unit. This method has lower complexity and power consumption than using SAR ADC. Therefore, PDM has gradually become a popular digital interface method for low-power MEMS sensors.

[0005] On the other hand, a disadvantage of PDM signals is that they cannot be directly digitally processed like pulse code modulation (PCM) signals. The computing units of modern digital circuit technology are all based on PCM signal operations, and there is no digital processing unit that directly supports PDM signals. Therefore, before using PDM signals for digital computing-based algorithm processing, they must be converted into amplitude-coded digital signals (such as PCM) with a certain bit width. This conversion process is called PDM decoding, which is generally completed through multi-stage digital filtering and multi-stage downsampling extraction. Not only does it require separate hardware resources to complete, but it also introduces corresponding processing delays. These delays cannot be compensated in any way, and the decoder will sacrifice the output signal quality in order to reduce processing delays when designing, and both cannot be achieved. Therefore, in many applications that require real-time processing of sensor signals, such as acoustic front-end processing algorithms in natural language processing based on microphone arrays, real-time estimation of direction of arrival (DOA) based on sensor arrays, and feedforward or feedback active noise reduction applications based on adaptive filter principles, PDM decoding delay has become a bottleneck that limits the performance of the above applications. A low-latency PDM signal processing method is needed to improve the current dilemma, which is also a technical challenge that the industry generally pays attention to. Summary of the invention

[0006] The purpose of the present invention is to solve the problem of reducing processing delay in PDM signal processing in the above-mentioned prior art without affecting the quality of the output signal, and to provide an adaptive filter based on a 1-bit digital signal and its application, which can directly perform adaptive filtering operations on the PDM signal without converting it to a multi-bit digital amplitude coded signal.

[0007] In order to achieve the above object, the present invention has the following technical solutions:

[0008] An adaptive filter based on 1-bit digital signal operation includes a digital signal processing operation module. The digital signal processing operation module uses the 1-bit digital stream signal to directly perform an adaptive filtering operation and then outputs it.

[0009] The present invention also provides an application of an adaptive filter based on a 1-bit digital signal for operation in adaptive system identification. The adaptive identification system consists of two parts, a digital domain and an analog domain. The digital domain part includes a reference signal generation module, a digital filter module, an adaptive coefficient update module and a bit width conversion module, and the analog domain part includes an unknown system to be identified and an error sensor; the reference signal generation module generates a 1-bit digital stream signal and sends it to the digital filter module, the unknown system and the adaptive coefficient update module; the error sensor collects the output of the unknown system and outputs it to the digital domain part in the form of a 1-bit digital stream signal, and after the bit width is converted to N bits, the N-bit PCM signal is subtracted from the N-bit PCM output signal output by the digital filter module to obtain an N-bit PCM signal and sent to the adaptive coefficient update module, and they jointly participate in the adjustment and iteration of the adaptive digital filter coefficients.

[0010] The 1-bit digital stream signal is a PDM signal.

[0011] The digital filter module performs convolution operation with L filter coefficients of N-bit length generated by the adaptive coefficient update module; N is the bit width of the sampling signal output by the decoder, N>1; L is the number of digital filter taps.

[0012] For LMS, the adaptive coefficient update module adjusts and iterates the filter coefficients according to the following formula:

[0013] Wj(n+1)=Wj(n)+mju*Xj(n)*err(n)

[0014] Among them, Wj(n) represents the value of the j-th tap coefficient of the filter at the n-th clock moment; Xj(n) represents the input data value on the j-th tap of the filter at the n-th clock moment, which is a 1-bit digital stream signal; err(n) represents the error signal value at the n-th clock moment, which is an N-bit PCM signal; mju is the adjustment step size, which is time-varying.

[0015] The present invention also provides an application of an adaptive filter based on a 1-bit digital signal for operation in Fx LMS adaptive signal cancellation. The adaptive cancellation system consists of two parts, a digital domain and an analog domain. The digital domain part includes a digital filter module, an adaptive coefficient update module and an Fx filter module, and the analog domain part includes a primary path system, a secondary path system and an error sensor. A 1-bit digital stream signal is generated by a reference sensor and sent to the digital filter module, the Fx filter module and the primary path system. The digital filter module generates an N-bit PCM signal and sends it to the secondary path system to generate an analog signal to cancel the analog signal generated by the primary path system. The analog signal output after cancellation generates a 1-bit digital stream signal through the error sensor and inputs the adaptive coefficient update module, which participates in the adjustment and iteration of the adaptive digital filter coefficient together with the N-bit PCM signal also output by the Fx filter module.

[0016] The present invention also provides an application of an adaptive filter based on a 1-bit digital signal for operation in Fe LMS adaptive signal cancellation. The adaptive cancellation system consists of two parts, a digital domain and an analog domain. The digital domain part includes a digital filter module, a delay module, an adaptive coefficient update module and a Fe filter module, and the analog domain part includes a primary path system, a secondary path system and an error sensor. A 1-bit digital stream signal is generated by a reference sensor and sent to the digital filter module, the delay module and the primary path system. The digital filter module generates an N-bit PCM signal and sends it to the secondary path system to generate an analog signal to cancel the analog signal generated by the primary path system. The analog signal output after cancellation generates a 1-bit digital stream signal through the error sensor and is input into the Fe filter module. The N-bit PCM signal generated by the Fe filter module and the 1-bit digital stream signal generated by the delay module are input into the adaptive coefficient update module to adjust and iterate the digital filter coefficient.

[0017] The delay time of the delay module is the filter order of the Fe filter module multiplied by the sampling period of the 1-bit digital stream signal.

[0018] Compared with the prior art, the present invention has at least the following beneficial effects: directly using a 1-bit digital stream signal as the input signal for digital adaptive filtering without converting the 1-bit digital stream signal into a multi-bit PCM signal in advance. The adaptive filtering algorithm is extended from PCM-based operations to 1-bit digital stream signals through a small modification, so that when using a digital signal operation unit to process a PDM signal generated by a sensor, additional PDM to PCM conversion hardware can be omitted, reducing complexity and processing delay. In the internal calculation of the digital filter module output calculation and the adaptive coefficient update module, the adaptive filter can omit all multipliers, and the complexity of the entire system will inevitably be greatly reduced. At the same time, the timing margin of the system clock speed will increase a lot, which has great positive significance for the reliability and stability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 Schematic diagram of the digital signal processing flow of traditional filters;

[0020] Figure 2 A schematic diagram of the processing flow of an adaptive filter based on a 1-bit digital signal for operation in the present invention;

[0021] Figure 3 The principle block diagram of the application of the adaptive filter of the present invention in adaptive system identification;

[0022] Figure 4 The principle block diagram of the application of the adaptive filter of the present invention in Fx LMS adaptive signal cancellation;

[0023] Figure 5 Principle block diagram of the application of the adaptive filter of the present invention in Fe LMS adaptive signal cancellation. DETAILED DESCRIPTION

[0024] The present invention is further described in detail below in conjunction with the accompanying drawings.

[0025] The widespread application of digital signal processing technology in various fields has brought unprecedented progress to these fields. Among them, adaptive filtering technology, as a representative of modern signal processing technology, has played an important role. Adaptive filtering technology is an optimal filtering method developed on the basis of linear filtering such as Wiener filtering and Kalman filtering. Because it has stronger dynamic adaptability and better filtering performance, it has been widely used in engineering practice, especially in information processing technology. Adaptive filtering is common in many different fields such as acoustics, speech, signal processing, control, image processing, etc. It is an intelligent and more targeted filtering method. Its typical application is for system identification and adaptive signal cancellation.

[0026] 1-bit digital stream level signal is a method of transmitting low-speed analog signals or low-speed multi-bit digital signals by high-speed 1-bit logic level jump. An analog or digital signal is represented by the frequency of logic level jump or the proportion of high level duration. PCM signal encodes the signal amplitude at the sampling moment and represents it with a group of bits containing multiple bits. The encoding method can choose different schemes according to the characteristics of the application system.

[0027] When a 1-bit digital stream level signal is used to transmit a low-speed multi-bit digital signal, a low-speed digital signal with N bits representing the amplitude of each sampling point is usually used as the input of a specific digital encoder, and the output of the encoder is a high-speed 1-bit digital stream signal, the spectrum of which contains the complete spectrum of the original N-bit digital signal and some additional high-frequency interference energy introduced by the encoder. (For example, the encoding method of using the occurrence density of a 1-bit pulse signal to represent the amplitude of an N-bit signal. However, the 1-bit digital stream signal described in the present invention is not limited to this encoding method.) Since the interference energy is in the high-frequency part and is separated from the spectrum of the signal in the spectrum, the 1-bit signal can be filtered out or suppressed by a simple analog filter to obtain the corresponding original analog signal.

[0028] A 1-bit digital stream level signal can also be a 0, 1 bit sequence converted from an analog signal through a specific modulator. In this case, the 1-bit digital stream signal is considered a digital signal. The benefit of this conversion is that it is conducive to long-distance transmission without causing any damage to the in-band signal spectrum shape, and only requires one data signal line and one clock signal line. In order to facilitate the subsequent digital signal processing, a specific decoder is also required to convert the high-speed 1-bit signal into a lower-speed N-bit digital signal. Therefore, the 1-bit digital stream signal can be regarded as a special digitization method for analog signals.

[0029] Since the PDM signal is an analog signal in digital form, the PDM signal has a natural advantage, which is that it can directly drive the digital power amplifier (Class D) chip without the need for a new digital-to-analog conversion chip (circuit), saving complexity while also reducing the delay in the signal path. If there is a digital operation processing method that can directly perform adaptive filtering operations on the PDM signal without converting to a multi-bit digital amplitude coded signal (such as PCM), then the PDM to PCM decoding delay and PCM to PDM encoding delay or DAC delay (such as Figure 1 This can greatly improve the real-time performance of the system and break through the performance bottleneck previously constrained by data path delay.

[0030] See also Figure 2The adaptive filter based on 1-bit digital signal operation of the present invention includes a digital signal processing operation module, which directly uses 1-bit digital stream signal to perform adaptive filtering operation and then output. In the traditional filter digital signal processing, the total delay caused by the PDM / PCM decoder and PCM / PDM encoder is τ1+τ2, while the PDM-based digital signal processing operation of the present invention can reduce the delay by τ1+τ2 compared with the traditional solution.

[0031] See also Figure 3 , Application of adaptive filter based on 1-bit digital signal in adaptive system identification: The adaptive identification system consists of two parts: digital domain and analog domain. The digital domain part includes reference signal generation module, digital filter module, adaptive coefficient update module and bit width conversion module, and the analog domain part includes unknown system to be identified and error sensor; the reference signal generation module generates 1-bit digital stream signal and sends it to digital filter module, unknown system and adaptive coefficient update module; the error sensor collects the output of unknown system and outputs it to digital domain part in the form of 1-bit digital stream signal, after bit width conversion to N bits, it is subtracted from the N-bit PCM output signal output by digital filter module to obtain N-bit PCM signal and sent to adaptive coefficient update module to participate in the adjustment and iteration of adaptive digital filter coefficients. 1-bit digital stream signal is PDM signal. In digital filter module, convolution operation is completed with L N-bit filter coefficients generated by adaptive coefficient update module; N is the bit width of sampling signal output by decoder, N>1; L is the number of digital filter taps.

[0032] For LMS, the adaptive coefficient update module adjusts and iterates the filter coefficients according to the following formula:

[0033] Wj(n+1)=Wj(n)+mju*Xj(n)*err(n)

[0034] Among them, Wj(n) represents the value of the j-th tap coefficient of the filter at the n-th clock moment; Xj(n) represents the input data value on the j-th tap of the filter at the n-th clock moment, which is a 1-bit digital stream signal; err(n) represents the error signal value at the n-th clock moment, which is an N-bit PCM signal; mju is the adjustment step size, which can be time-varying.

[0035] As can be seen from the formula, the adjustment amount of the tap coefficient is obtained by multiplying the step length by the product of the input data value Xj(n) (1-bit PDM signal) on the j-th tap of the filter and the error signal value err(n) (N-bit PCM) corresponding to the same time. Assuming that the step length is an integer power of 2, the two multiplication operations in the tap coefficient update are completely eliminated in the current case. This is because the Xj(n) involved in the multiplication is 1-bit PDM. Similarly, the convolution operation of the input signal (1-bit PDM) and the filter coefficient in the digital filter module can also be completed without a multiplier, which greatly reduces the complexity of the hardware implementation, because the multiplier is a high-complexity and high-delay operation unit in the entire digital circuit, and is also the main bottleneck limiting the system clock speed. Once the multiplier unit is no longer needed, the complexity of the entire system will inevitably be greatly reduced, and the timing margin of the system clock speed will increase a lot, which has great positive significance for the reliability and stability of the system.

[0036] Under given initial conditions (including coefficient iteration step, simulated transmission path, etc.), the system can quickly converge the filter coefficients to a stable value, which makes the error signal err(n) reach the minimum under the MSE criterion. The transfer function composed of the filter coefficients at this time is the best estimate of the unknown system H(z) under the MSE criterion.

[0037] See also Figure 4 , an application of an adaptive filter based on 1-bit digital signal operation in Fx LMS adaptive signal cancellation. Taking the LMS algorithm as an example, the adaptive cancellation system consists of two parts: digital domain and analog domain. The digital domain part includes a digital filter module, an adaptive coefficient update module and an Fx filter module, and the analog domain part includes a primary path system, a secondary path system and an error sensor. A 1-bit digital stream signal generated by a reference sensor is sent to the digital filter module, the Fx filter module and the primary path system. The digital filter module generates an N-bit PCM signal and sends it to the secondary path system to generate an analog signal to cancel the analog signal generated by the primary path system. The analog signal output after cancellation generates a 1-bit digital stream signal through the error sensor and inputs the adaptive coefficient update module. Together with the N-bit PCM signal also output by the Fx filter module, it participates in the adjustment and iteration of the adaptive digital filter coefficient. The primary path system has a transfer function of the 1-bit digital signal generated by the reference sensor to the error sensor through analog transmission. The secondary path system has an analog path transfer function of the N-bit digital PCM signal output by the digital filter to the error sensor through analog transmission.

[0038] See also Figure 5, an application of an adaptive filter based on 1-bit digital signal operation in Fe LMS adaptive signal cancellation. Taking LMS algorithm as an example, the adaptive cancellation system consists of two parts: digital domain and analog domain. The digital domain part includes digital filter module, delay module, adaptive coefficient update module and Fe filter module, and the analog domain part includes primary path system, secondary path system and error sensor. The reference sensor generates a 1-bit digital stream signal and sends it to the digital filter module, delay module and primary path system. The digital filter module generates an N-bit PCM signal and sends it to the secondary path system to generate an analog signal to cancel the analog signal generated by the primary path system. The analog signal after cancellation generates a 1-bit digital stream signal through the error sensor and inputs the Fe filter module. The N-bit PCM signal generated by the Fe filter module and the 1-bit digital stream signal generated by the delay module are input into the adaptive coefficient update module to adjust and iterate the digital filter coefficient. The delay time of the delay module is the filter order of the Fe filter module multiplied by the sampling period of the 1-bit digital stream signal.

[0039] The Fx filter is a finite impulse response digital filter, which is an equivalent approximation of the "secondary path system" in the sense of minimizing the minimum mean square error. The Fe filter module is a finite impulse response digital filter obtained by reversing the tap coefficients of the Fx filter. Whether it is the Fx LMS algorithm or the Fe LMS algorithm, the multiplier in the operation in the digital domain can be omitted, because one of the multiplier and the multiplicand in the multiplication operation in the original formula is always a 1-bit PDM signal.

[0040] The present invention uses an internally generated 1-bit PDM signal or a 1-bit PDM signal generated from an external sensor as a data input or / and error input of a fully digital adaptive filter and directly uses these signals to perform all operations required for adaptive filtering without prior conversion of PDM to PCM signals. The adaptive coefficient adjustment module in the adaptive filter always ensures that at least one of the input signals of all multiplication operations is a 1-bit PDM signal or a multi-bit PCM signal equivalent to 1 bit. The present invention reduces complexity, reduces processing delay, and improves reliability.

[0041] The above description is only a preferred embodiment of the present invention and is not intended to impose any limitation on the technical solution of the present invention. Those skilled in the art should understand that, without departing from the spirit and principles of the present invention, the technical solution can also be subjected to several simple modifications and substitutions, and these modifications and substitutions are also within the scope of protection covered by the claims.

Claims

1. An application of an adaptive filter based on a 1-bit digital signal for adaptive system identification, characterized in that: The adaptive filter based on 1-bit digital signal operation includes a digital signal processing operation module, which uses a 1-bit digital stream signal to directly perform adaptive filtering operation and then output; the adaptive identification system consists of two parts, a digital domain and an analog domain, the digital domain part includes a reference signal generation module, a digital filter module, an adaptive coefficient update module and a bit width conversion module, and the analog domain part includes an unknown system to be identified and an error sensor; the reference signal generation module generates a 1-bit digital stream signal and sends it to the digital filter module, the unknown system and the adaptive coefficient update module; The error sensor collects the output of the unknown system and outputs it to the digital domain in the form of a 1-bit digital stream signal. After the bit width is converted to N bits, it is subtracted from the N-bit PCM output signal output by the digital filter module to obtain an N-bit PCM signal that is sent to the adaptive coefficient update module to jointly participate in the adjustment and iteration of the adaptive digital filter coefficients.

2. The application of the adaptive filter based on 1-bit digital signal in adaptive system identification according to claim 1, characterized in that: The 1-bit digital stream signal is a PDM signal.

3. The application of the adaptive filter based on 1-bit digital signal operation in adaptive system identification according to claim 1 is characterized in that: The digital filter module performs convolution operation with L filter coefficients of N-bit length generated by the adaptive coefficient update module; N is the bit width of the sampling signal output by the decoder, N>1; L is the number of digital filter taps.

4. The application of the adaptive filter based on 1-bit digital signal operation in adaptive system identification according to claim 1 is characterized in that: For LMS, the adaptive coefficient update module adjusts and iterates the filter coefficients according to the following formula: Wj(n+1)=Wj(n)+mju*Xj(n)*err(n) Among them, Wj(n) represents the value of the j-th tap coefficient of the filter at the n-th clock moment; Xj(n) represents the input data value on the j-th tap of the filter at the n-th clock moment, which is a 1-bit digital stream signal; err(n) represents the error signal value at the n-th clock moment, which is an N-bit PCM signal; mju is the adjustment step size, which is time-varying.

5. An application of an adaptive filter based on a 1-bit digital signal for Fx LMS adaptive signal cancellation, characterized in that: The adaptive filter based on 1-bit digital signal operation includes a digital signal processing operation module, which uses a 1-bit digital stream signal to directly perform adaptive filtering operation and then output; the adaptive cancellation system consists of two parts, a digital domain and an analog domain, the digital domain part includes a digital filter module, an adaptive coefficient update module and an Fx filter module, and the analog domain part includes a primary path system, a secondary path system and an error sensor; a 1-bit digital stream signal generated by a reference sensor is sent to the digital filter module, the Fx filter module and the primary path system, the digital filter module generates an N-bit PCM signal and sends it to the secondary path system to generate an analog signal to offset the analog signal generated by the primary path system, and the analog signal output after the offset generates a 1-bit digital stream signal through the error sensor and inputs the adaptive coefficient update module, and participates in the adjustment and iteration of the adaptive digital filter coefficient together with the N-bit PCM signal also output by the Fx filter module.

6. An application of an adaptive filter based on a 1-bit digital signal for Fe LMS adaptive signal cancellation, characterized in that: The adaptive filter based on 1-bit digital signal operation includes a digital signal processing operation module, which uses a 1-bit digital stream signal to directly perform adaptive filtering operation and then output; the adaptive cancellation system consists of two parts, a digital domain and an analog domain, the digital domain part includes a digital filter module, a delay module, an adaptive coefficient update module and an Fe filter module, and the analog domain part includes a primary path system, a secondary path system and an error sensor; a 1-bit digital stream signal generated by a reference sensor is sent to the digital filter module, the delay module and the primary path system, the digital filter module generates an N-bit PCM signal and sends it to the secondary path system to generate an analog signal to offset the analog signal generated by the primary path system, the analog signal output after the offset generates a 1-bit digital stream signal through the error sensor and inputs the Fe filter module, the N-bit PCM signal generated by the Fe filter module and the 1-bit digital stream signal generated by the delay module are input into the adaptive coefficient update module to adjust and iterate the digital filter coefficient.

7. The application of the adaptive filter based on 1-bit digital signal operation in Fe LMS adaptive signal cancellation according to claim 6, characterized in that: The delay time of the delay module is the filter order of the Fe filter module multiplied by the sampling period of the 1-bit digital stream signal.

Images