Zynq-based voice data acquisition and processing device

Abstract

The invention relates to a voice data acquisition and processing device based on Zynq, and the device comprises a voice acquisition module which is used for synchronously acquiring an environment sound signal in a machine and a voice signal of a main loudspeaker, and converting the environment sound signal and the voice signal into a digital voice signal and a digital environment sound signal for output after filtering; the noise estimation module is used for carrying out noise estimation according to the input digital voice signal and outputting a noise estimation result; the noise reduction module is used for carrying out noise reduction processing on the digital voice signal according to the noise estimation result to obtain a pure voice signal; the sound mixing module is used for mixing the pure voice signal and the digital environment sound signal to obtain a sound mixing signal; the silencing module is used for shielding noise input when no voice is played; the noise estimation module, the noise reduction module, the sound mixing module and the silencing module are integrated on a single Zynq SOC. The design and development cost of the built-in communication system can be reduced, the power consumption of a product is reduced, and the practicability of the system is improved.

Description

Technical Field

[0001] This invention relates to the field of in-flight communication systems, and in particular to a voice data acquisition and processing device based on Zynq. Background Technology

[0002] The in-flight intercom system is a device that enables communication between passengers inside the aircraft and between passengers inside the aircraft and external personnel. The in-flight intercom system consists of an in-flight intercom control box, headset microphones, a PTT (Push-to-Talk) system, and cables. When passengers inside the aircraft communicate, the in-flight intercom control box collects the voice signal, performs noise reduction processing, and then uses mixing and bus communication technologies to play the processed voice signal through the headset. Due to the high noise level, poor electronic and electrical environment, and complex logic caused by the variety of hinged devices in aircraft environments, voice acquisition and voice processing are the key technologies of the in-flight intercom system.

[0003] In-flight communication systems typically employ a dual-chip architecture such as FPGA+DSP or FPGA+PPC for data acquisition and processing. The FPGA handles voice signal acquisition and control, while the processor chip runs algorithms for noise reduction and mixing, and implements interface communication functions. This dual-chip architecture leads to complex hardware design, high power consumption, and cumbersome software development, resulting in high design and implementation costs. Furthermore, the two chips in a dual-chip architecture interact via a circuit board, exposing critical data to the complex electromagnetic environment within the device. This introduces interference and can cause signal processing delays and synchronization issues. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a voice data acquisition and processing device based on Zynq, which can reduce the design and development cost of the in-flight communication system, reduce product power consumption, and increase the practicality of the system.

[0005] The technical solution adopted by this invention to solve its technical problem is: to provide a voice data acquisition and processing device based on Zynq, comprising:

[0006] The voice acquisition module is used to simultaneously acquire ambient sound signals and voice signals from the main speaker, and then filter and convert them into digital voice signals and digital ambient sound signals for output.

[0007] The noise estimation module is used to perform noise estimation based on the input digital voice signal and output the noise estimation result.

[0008] The noise reduction module is used to process the digital voice signal for noise reduction based on the noise estimation results, so as to obtain a clean voice signal;

[0009] The mixing module is used to mix the clean voice signal with the digital ambient sound signal to obtain the mixed signal;

[0010] The noise suppression module is used to block noise input when there is no voice playback.

[0011] The noise estimation module, noise reduction module, mixing module, and squelch module are integrated on a single Zynq SOC.

[0012] Furthermore, the noise estimation results include the signal-to-noise ratio and spectral characteristics of the digital voice signal.

[0013] Furthermore, it also includes a voice enhancement module, which adjusts the matching degree of voice energy and frequency according to the spectral characteristics of the digital voice signal.

[0014] Furthermore, the voice acquisition module includes a filtering module and an AD / DA module.

[0015] Furthermore, the filtering parameters of the filtering module are adjusted based on the noise estimation results.

[0016] Furthermore, the sampling rate and sampling accuracy of the AD section in the AD / DA module are digitally adjusted according to the sampling requirements of the voice signal in different scenarios.

[0017] Furthermore, the ambient sound signals include multiple audio signals from emergency intercom, emergency radio, face mask, and attendants. These signals are selected and gated through mode selection logic, and the corresponding audio signals are output to the Zynq SOC.

[0018] Furthermore, the noise estimation module is deployed on the PL side of the Zynq SOC.

[0019] Furthermore, the algorithm logic of the noise reduction module is deployed on the PS side of the Zynq SOC and is hardware accelerated through the PL side of the Zynq SOC and the DSP48E1 module.

[0020] Beneficial effects

[0021] Due to the adoption of the above technical solutions, this invention has the following advantages and positive effects compared with the prior art: This invention uses adjustable analog filtering combined with AD signal conditioning and adjustable digital sampling to ensure the accuracy of the acquired digital signal, laying a high signal-to-noise ratio foundation for subsequent noise processing and speech enhancement; This invention deploys computationally intensive tasks on the PL end of the Zynq SOC and algorithm logic tasks on the PS end, and uses a DSP48E1 digital signal processing unit for hardware acceleration. It utilizes the parallel processing characteristics of the PL end to solve the problems of high latency and poor real-time performance of pure software processing, while using the algorithm flexibility of the PS end to make up for the weak adaptability of pure hardware processing scenarios. Attached Figure Description

[0022] Figure 1 This is a system architecture diagram of an embodiment of the present invention;

[0023] Figure 2 This is a schematic diagram of the adaptive voice data acquisition structure according to an embodiment of the present invention;

[0024] Figure 3 This is a schematic diagram of the voice signal processing structure according to an embodiment of the present invention. Detailed Implementation

[0025] The present invention will be further illustrated below with reference to specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. Furthermore, it should be understood that after reading the teachings of this invention, those skilled in the art can make various alterations or modifications to the invention, and these equivalent forms also fall within the scope defined by the appended claims.

[0026] The embodiments of the present invention relate to a voice data acquisition and processing device based on Zynq, which includes several parts such as adaptive voice data acquisition, voice data noise reduction processing, mixing and squelching. Through SOC chip and modular peripheral circuits, a general architecture for voice data acquisition and processing applied to in-flight communication control system is realized.

[0027] This implementation can realize the entire process including voice signal acquisition and processing, and the functions of the in-set call control box, such as acquisition input, analog-to-digital conversion, algorithm processing, and digital-to-analog conversion output, for noisy voice.

[0028] The noisy voice signal is initially filtered out by adaptive voice data acquisition. The input signal is converted from digital to analog by an AD module with pre-set sampling parameters for subsequent digital signal processing. The acquired digital signal is analyzed for noise and voice features by a Zynq chip, and noise reduction is performed based on the analysis results. The processed audio is mixed with other sound information and normalized, and then converted into an analog signal by a DA module for playback.

[0029] like Figure 1 As shown, this is a preferred embodiment 1 of the present invention, which includes main modules such as power supply, adaptive voice data acquisition, and ZynqSOC.

[0030] The adaptive voice data acquisition module includes a parameter-adjustable filtering module and an AD / DA conversion module. The parameter-adjustable filtering module adaptively adjusts the filtering coefficients based on estimated parameters of the voice and noise information to filter out significant noise and improve the signal-to-noise ratio of the voice signal. The AD / DA conversion module converts the acquired raw analog voice signal into a digital signal; and converts the processed and synthesized digital voice signal back into an analog signal.

[0031] More specifically, the parameter-adjustable filtering module includes an analog low-pass filter to initially filter out high-frequency noise, adapt to the sampling rate, and avoid aliasing. The analog low-pass filter receives parameter configuration from the PL terminal via SPI to achieve initial high-frequency noise filtering and adaptive filtering to different noise environments within the machine.

[0032] The AD / DA conversion module is used to perform AD / DA conversion and can also integrate analog noise reduction circuitry to supplement the shortcomings of digital noise reduction. The AD / DA conversion module includes a signal conditioning section to adapt to the input signal voltage parameters of the AD section and the matching of the output signal of the DA section with the headphones. Furthermore, the sampling rate and sampling accuracy of the AD section can be digitally adjusted to match the sampling requirements of the voice signal in different scenarios.

[0033] The conditioning / logic can control the connection to analog acquisition channels such as "emergency intercom" and "emergency radio" to achieve signal amplification, filtering and other preprocessing to ensure the signal quality of input Zynq.

[0034] The Zynq SoC consists of a PS (Programmable Logic Unit) and a PL (Programmable Logic Unit): the PL implements programmable hardware logic and is responsible for real-time hardware tasks; the PS handles algorithms and system control. The PS and PL communicate at high speed via the LBC (Low-Cost Per Module) bus. The PS connects to DDR, FLASH, and EEPROM via LBC, and transmits data at high speed to the bus board via PCIe.

[0035] In this embodiment, all voice data noise reduction processing is implemented on a single Zynq SOC:

[0036] The noise estimation and analysis module is implemented in hardware language on the Zynq PL end. It can quickly estimate the acquired raw speech and noise signals, evaluate the signal-to-noise ratio, analyze the spectrum of the raw speech and noise, and extract feature parameters. In addition to signal-to-noise ratio and signal spectrum, a prediction algorithm can also be used to make a preliminary prediction of the noise mixed in the next acquisition in the time domain, which is convenient for parameter adjustment.

[0037] The noise reduction module is implemented through the Zynq PL and PS terminals. The PS terminal is responsible for processing the algorithm logic, while the PL terminal and DSP48E1 module perform hardware acceleration to reduce processing latency and perform adaptive noise reduction processing on the input signal to extract the voice signal. The noise reduction depth can be digitally adjusted to adapt to the usage requirements of different noise environments.

[0038] The speech enhancement module receives the speech signal from the noise reduction module, combines it with the speech frequency information, and adjusts the matching degree between speech energy and frequency through normalization processing.

[0039] The mixing module will mix the collected voice signals, voice signals sent by other in-unit intercom control, radio data, alarm information, etc., through the mixing algorithm so that they can be played on the same timeline as required.

[0040] The squelch module assesses noise intensity based on estimated parameters of voice and noise information, sets a threshold, and blocks noise input when there is no voice playback.

[0041] like Figure 2 As shown, this is a preferred embodiment 2 of the present invention, which provides a specific architecture for the adaptive voice data acquisition part based on embodiment 1.

[0042] The voice data is multi-source voice input, covering multi-user, multi-device scenarios for in-phone calls (including microphones, oxygen masks, passive follower inputs, active follower inputs, passive follower outputs, active follower outputs, headsets, etc.). The voice data acquisition section includes a parameter-adjustable filtering module and an AD / DA conversion module, realizing preliminary noise filtering, digital-to-analog / analog-to-digital conversion, and signal amplification and conditioning functions.

[0043] The parameter-adjustable filtering module consists of a low-pass filter circuit and a parameter-adjustable filter. The low-pass filter circuit filters out high-frequency noise, and the parameter-adjustable filter performs adaptive filtering to adapt to the working environment and usage scenario.

[0044] AD / DA conversion includes a digital-to-analog / analog-to-digital converter chip with adjustable parameters and a signal conditioning section. It can digitally set sampling parameters, including sampling rate and sampling accuracy, to achieve functions such as signal amplification, smoothing, and digital-to-analog / analog-to-digital conversion.

[0045] Signal conditioning is the core of adaptive acquisition. Each input source corresponds to an independent input conditioning circuit to address the differences between multiple signal sources. Mode selection logic serves as the core of input source adaptation, switching input channels and adjusting gain / filter parameters according to system configuration. Input and output conditioning amplifiers are used to adjust signal amplitude to match the input range of the noise reduction / AD / DA circuits. The noise reduction / AD / DA circuits interface with the conditioned analog signal to achieve analog noise reduction and analog-to-digital / digital-to-analog conversion.

[0046] In addition, this embodiment also incorporates a BIT test circuit, which generates a test signal that is injected into the conditioning link to achieve system self-testing and ensure the reliability of the analog front end.

[0047] like Figure 3 As shown, this is a preferred embodiment 3 of the present invention, which provides a specific architecture for the voice signal processing part based on embodiment 2.

[0048] Voice signal processing includes voice data noise reduction, mixing, and squelching, which are used to realize the digital processing functions of voice signals.

[0049] Voice data denoising processing includes noise estimation and analysis, a denoising module, and speech enhancement. Noise estimation and analysis performs functions such as signal-to-noise ratio estimation, spectrum analysis, and noise prediction for both voice and noise, providing processing parameters for adaptive voice acquisition and denoising enhancement. The denoising module incorporates a voice denoising algorithm that uses voice noise analysis parameters as input to adaptively process noise in noisy voice. The speech enhancement module uses voice noise analysis parameters as input to adjust the matching degree between the frequency and intensity of the processed voice signal, further improving voice quality.

[0050] Mixing and squelching are divided into mixing and squelching functions. The mixing function normalizes the enhanced voice signal and sound information from different sources before mixing them to form a time-domain audio signal containing multiple sources. The squelching function adaptively sets the threshold based on voice noise analysis parameters to block noise when there is no voice playback, thus achieving a comfortable user experience for the in-set intercom control box in different usage environments.

Claims

1. A voice data acquisition processing device based on Zynq, characterized in that, The application relates to a voice processing system based on Zynq SOC, which comprises the following parts: a voice collection module, which is used for synchronously collecting an in-machine environment sound signal and a voice signal of a main loudspeaker, filtering and converting the signals into digital voice signals and digital environment sound signals for output; a noise estimation module, which is used for noise estimation according to the input digital voice signals and outputs noise estimation results; a noise reduction module, which is used for noise reduction processing of the digital voice signals according to the noise estimation results and obtains pure voice signals; a mix sound module, which is used for mixing the pure voice signals with the digital environment sound signals to obtain mixed sound signals; a noise elimination module, which is used for shielding noise input when there is no voice playing; the noise estimation module, the noise reduction module, the mix sound module and the noise elimination module are integrated on a single Zynq SOC.

2. The apparatus of claim 1, wherein, The noise estimation results include the signal-to-noise ratio and spectral characteristics of the digital voice signals.

3. The apparatus of claim 2, wherein, The voice processing system further comprises a voice enhancement module, which is used for adjusting the matching degree of voice energy and frequency according to the spectral characteristics of the digital voice signals.

4. The apparatus of claim 1, wherein, The voice collection module comprises a filtering module and an AD / DA module.

5. The apparatus of claim 4, wherein, The filtering parameters of the filtering module are adjusted according to the noise estimation results.

6. The apparatus of claim 4, wherein, The sampling rate and sampling precision of the AD part of the AD / DA module are digitally adjusted according to the sampling requirements of the voice signals in different scenes.

7. The apparatus of claim 1, wherein, The environment sound signals include multi-path sound signals from emergency intercoms, emergency radios, masks and attendants, and the corresponding sound signals are output to the Zynq SOC through mode selection logic.

8. The apparatus of claim 1, wherein, The noise estimation module is arranged on the PL end of the Zynq SOC.

9. The apparatus of claim 1, wherein, The algorithm logic of the noise reduction module is arranged on the PS end of the Zynq SOC and is accelerated by the PL end and the DSP48E1 module of the Zynq SOC.

Images