Voice signal silencing device based on joint detection

Abstract

The invention discloses a voice signal silencing device based on joint detection. A received intermediate frequency voice signal is subjected to analog-to-digital conversion and digital down-conversion through an AD module and a down-conversion module; after digital automatic gain control is completed by the AGC module, the voice signal is divided into two paths; the carrier detection processing module carries out carrier detection processing on one path of voice signals to obtain power spectrum characteristics; the voice squelch detection module demodulates the other path of voice signal, on one hand, demodulated voice is output when the squelch gate is in an open state, and on the other hand, voice feature detection processing is performed on the demodulated voice to obtain voice features; and the joint judgment module performs joint judgment on the power spectrum characteristics and the voice characteristics calculated by the carrier wave squelch detection module and the voice squelch detection module, and controls the opening and closing states of a squelch door. According to the invention, the carrier squelch technology and the voice squelch technology are jointly used, so that the false alarm probability and the missed alarm probability are effectively reduced.

Description

Technical Field

[0001] This invention belongs to the field of wireless voice communication technology and relates to a voice signal squelching device based on joint detection. Background Technology

[0002] In voice communication, background noise during radio reception (when there is no voice transmission) can affect the user's auditory experience. Squelch technology can automatically turn the handset on or off depending on whether there is voice transmission, thus protecting the user from noise. Currently, there are three main types of squelch technology: carrier squelch, pilot squelch, and voice feature squelch.

[0003] Carrier squelch technology determines the presence of voice signal transmission by detecting the presence of a carrier in the intermediate frequency signal. However, it is susceptible to misjudgment due to single-carrier interference. Furthermore, for AM modulation, a common mode for voice communication, the carrier power of AM signals with different modulation intensities fluctuates. Under the same signal-to-noise ratio (SNR), the calculated carrier-to-noise ratios for AM signals with different modulation intensities will differ. This means that a single threshold is not effective in squelch threshold determination, necessitating the introduction of additional modulation intensity identification operations or the use of more complex threshold decision methods. Both of these methods increase the resource consumption for squelch operation implementation.

[0004] Pilot squelch technology detects voice signals by adding an extra single tone as a marker. This technique requires inserting a low-frequency signal into the voice signal at the transmitting end and then demodulating and removing the pilot tone at the receiving end. Because the frequency distribution range of voice signals is very narrow, between 300Hz and 3.5kHz, the inserted single-tone pilot tone is often in the 0-300Hz band. Demodulating the voice requires filtering it using a bandpass filter with a very narrow transition band, which typically results in excessive resource overhead and affects voice quality.

[0005] Voice feature squelch involves processing the demodulated signal and identifying speech features. Common methods include speech spectrum recognition or short-time feature recognition. However, this method is highly susceptible to signal-to-noise ratio (SNR). At low SNR levels, the accuracy of voice feature squelch methods drops significantly, making them unsuitable for low SNR conditions.

[0006] In summary, voice squelch is an important part of voice communication. The process can be briefly described as calculating several feature values ​​and making a decision based on a set threshold. The calculation of these feature values ​​requires comprehensive consideration of recognition accuracy and resource overhead, while the squelch decision-making process needs to avoid frequent on / off switching of the voice output, necessitating a robust decision-making procedure. Summary of the Invention

[0007] The purpose of this invention is to provide a voice signal squelching device based on joint detection, which uses carrier squelching technology and voice feature squelching technology in combination, avoiding additional interference caused by the frequent opening and closing of the squelching decision switch.

[0008] The objective of this invention is achieved through the following technical solution:

[0009] A voice signal squelch device based on joint detection includes an AD module, a down-conversion module, an AGC module, a carrier squelch detection module, a voice squelch detection module, and a joint detection module;

[0010] The received intermediate frequency (IF) voice signal first undergoes analog-to-digital (ADC) and digital-to-digital (D / D) conversion via an AD module and a down-conversion module. Then, after digital automatic gain control via an AGC module, the voice signal is split into two paths. A carrier detection processing module directly performs carrier detection processing on one path to obtain power spectrum characteristics. A voice squelch detection module performs AM / FM demodulation on the other path, outputting the demodulated voice with the squelch gate open, and simultaneously performing voice feature detection processing on the demodulated voice to obtain voice features. A joint decision module performs a joint decision on the power spectrum characteristics and voice features calculated by the carrier squelch detection module and the voice squelch detection module to control the opening and closing state of the squelch gate.

[0011] Preferably, if the modulation mode of the carrier detection processing module is FM, the signal power within the signal bandwidth is selected as the main signal power; and the signal power deviating from the signal bandwidth by about 50% is selected as the noise floor power.

[0012] The ratio of the main signal power to the noise floor power is used as the power spectrum characteristic.

[0013] Preferably, if the modulation method of the carrier detection processing module is AM, the signal covered by the highest point of the power spectrum is taken as the main signal power, and a signal that deviates from the signal bandwidth by about 50% is selected as the noise floor power according to the same bandwidth range.

[0014] The difference between the main signal power and the noise floor power is used as a power spectrum feature.

[0015] Preferably, the voice squelch detection module first performs voice processing on the voice signal, including demodulation, DC removal, and filtering; then, the voice signal after voice processing is processed into frames, and the frames are input to two bandpass filters with the same gain. One bandpass filter outputs a high-frequency signal, and the other bandpass filter outputs a low-frequency signal. The square of the modulus of the low-frequency signal and the high-frequency signal is used as the time-domain energy value, and the ratio is processed to use as the voice feature.

[0016] Preferably, the joint decision based on power spectrum features and speech features includes the following steps:

[0017] Step 1: Squelch Decision Based on Power Spectrum Characteristics. First, the current squelch level is locked, and the carrier opening threshold and carrier closing threshold are determined. Then, the power spectrum characteristics are accumulated and averaged over N1 frames, and the average value is used as the power spectrum characteristic parameter. When the squelch gate is open, the power spectrum characteristic parameter is compared with the carrier closing threshold. If the power spectrum characteristic parameter is less than the carrier closing threshold, the carrier squelch result is output as 0; if the power spectrum characteristic parameter is greater than the carrier closing threshold, the carrier squelch result is output as 1. When the squelch gate is closed, the power spectrum characteristic parameter is compared with the carrier opening threshold. If the power spectrum characteristic parameter is greater than the carrier opening threshold, the carrier squelch result is output as 1; if the power spectrum characteristic parameter is less than the carrier opening threshold, the carrier squelch result is output as 0.

[0018] Step 2: Squelch determination based on voice features. First, the current squelch level is locked, and the voice opening threshold and voice closing threshold are determined. Then, voice features of N2 frames are accumulated, and the largest voice feature is used as the voice feature parameter. When the squelch gate is open, the voice feature parameter is compared with the voice closing threshold. If the voice feature parameter is less than the voice closing threshold, the voice squelch result is output as 0; if the voice feature parameter is greater than the voice closing threshold, the voice squelch result is output as 1. When the squelch gate is closed, the voice feature parameter is compared with the voice opening threshold. If the voice feature parameter is greater than the voice opening threshold, the voice squelch result is output as 1; if the voice feature parameter is less than the voice opening threshold, the voice squelch result is output as 0.

[0019] Step 3: Joint decision. If both the carrier squelch result and the voice squelch result are 0 and remain stable during the observation period, the squelch gate is closed. If it is unstable, the squelch gate is left as is. If both the carrier squelch result and the voice squelch result are 1, the squelch gate can be directly set to the open state. In other cases, the squelch gate is left as is.

[0020] The beneficial effects of this invention are as follows:

[0021] This invention enables noise reduction processing for common voice modulation schemes, AM and FM signals. By combining carrier noise reduction technology and voice noise reduction technology, the probability of false alarms and missed alarms is effectively reduced. Furthermore, addressing the issue of overly complex decision-making processes for AM modulated signals with different modulation depths in carrier noise detection, a front-end digital AGC is used to control the carrier power of AM signals with different modulation depths within the required range, simplifying the decision-making process. Attached Figure Description

[0022] Figure 1 This is a structural block diagram of a voice signal squelch device based on joint detection.

[0023] Figure 2This is a schematic diagram of the FM power spectrum.

[0024] Figure 3 This is a schematic diagram of the AM power spectrum.

[0025] Figure 4 This is a schematic diagram of the noise decision process based on power spectrum characteristics.

[0026] Figure 5 This is a schematic diagram of the time and frequency domains of a voice signal.

[0027] Figure 6 This is a schematic diagram of the noise reduction decision process based on speech features.

[0028] Figure 7 This is a schematic diagram of the joint judgment process. Detailed Implementation

[0029] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.

[0030] Voice squelch technology is crucial for voice communication, directly impacting its quality. This is especially true for aviation communications, where prolonged periods of ambient noise can severely impair pilot judgment. Furthermore, common voice signals often include brief pauses, and the squelch decision switch cannot be frequently switched on and off, causing additional interference. This embodiment addresses these issues by proposing a voice signal squelch device based on joint detection. This device combines carrier squelch technology and voice feature squelch technology, and optimizes the computation process, achieving effective squelch with minimal logic resources.

[0031] like Figure 1 As shown in the figure, this embodiment illustrates a voice signal squelch device based on joint detection, which includes an AD module, a down-conversion module, an AGC module, a carrier squelch detection module, a voice squelch detection module, and a joint detection module.

[0032] The received intermediate frequency (IF) voice signal first undergoes analog-to-digital (A / D) conversion and digital down-conversion via an AD module and a down-conversion module. Then, after automatic gain control (AGC) via an AGC module, the voice signal is split into two for signal processing. One signal goes directly to the carrier detection module for carrier detection processing to obtain power spectrum characteristics. The other signal goes to the voice squelch detection module for AM / FM (Amplitude Modulation / Frequency Modulation) demodulation. On one hand, the demodulated voice is output with the squelch gate open; on the other hand, voice feature detection processing is performed on the demodulated voice to obtain voice features. The modulation scheme of the received signal is known: AM is amplitude modulation, where the transmitted information controls the amplitude of the carrier wave; FM is frequency modulation, where the transmitted information controls the frequency of the carrier wave. For ease of description, the squelch gate's open / closed state is consistent with the voice output state; that is, the voice output open state is called the squelch gate open state, and the voice output closed state is called the squelch gate closed state. The joint decision module performs a joint decision on the power spectrum features and voice features calculated by the carrier squelch detection module and the voice squelch detection module, and controls the opening and closing state of the squelch gate.

[0033] The carrier detection processing module calculates the main signal power and the noise floor power according to different modulation methods, and uses the ratio of the main signal power to the noise floor power as the power spectrum characteristic value.

[0034] like Figure 2 The FM power spectrum shown indicates that if the current modulation mode is FM, the signal power within the signal bandwidth is selected as the main signal power; the signal power deviating from the signal bandwidth by about 50% is selected as the noise floor power to avoid interference caused by spectrum leakage; the noise floor power is calculated through the power spectrum.

[0035] Main signal power Calculate using the following formula:

[0036]

[0037] Among them, B FM-m Represents the main signal bandwidth of the FM signal; PSD FM (ω) represents the power spectrum of the FM signal, and its amplitude is expressed in decibels.

[0038] Noise floor power Calculate using the following formula:

[0039]

[0040] Among them, B FM-ln B represents the noise floor bandwidth to the left of the main signal;FM-rn This represents the noise floor bandwidth to the right of the main signal; both bandwidths are half the bandwidth of the main signal.

[0041] The power of the main signal is divided by the power of the noise floor, and the ratio is used as the power spectrum feature. The power spectrum feature values ​​of N1 frames are accumulated and averaged to serve as the final decision parameter.

[0042] like Figure 3 The AM power spectrum shown shows that if the current modulation mode is AM, the power of AM voice signals with different modulation levels will be controlled to a relatively consistent level due to the use of AGC technology at the front end. When calculating the main signal power, only the signal covered by the highest point of the power spectrum needs to be calculated as the main signal power, and the signal that deviates from the signal bandwidth by about 50% is selected as the noise floor power according to the same bandwidth range.

[0043] Main signal power Calculate using the following formula:

[0044]

[0045] Where, ω max N represents the frequency corresponding to the highest point of the power spectrum. AM This represents the data points covered by a 3dB attenuation at the highest point; PSD AM (ω) represents the power spectrum of the AM voice signal.

[0046] Noise floor power Calculate using the following formula:

[0047]

[0048] Where, ω AM_n This indicates a frequency point that is 5 kHz off the AM signal bandwidth.

[0049] The difference between the main signal power and the noise floor power is used as the power spectrum feature. The average value of the accumulated power spectrum feature values ​​of N1 frames is used as the final parameter.

[0050] Through this calculation, the different modulation levels will not affect the change in the main signal power. Therefore, there is no need to perform additional modulation level identification processing on AM signals with different modulation levels, or to use complex dynamic thresholds for discrimination.

[0051] Squelch level is selected by the user based on actual usage conditions. A higher squelch level results in less audible noise and a higher threshold value, suitable for environments with a high signal-to-noise ratio. Conversely, a lower squelch level results in more audible noise and a lower threshold value, suitable for environments with a low signal-to-noise ratio. The squelch decision process based on power spectrum characteristics is as follows: Figure 4As shown, the current squelch level needs to be locked first, and the carrier opening threshold and carrier closing threshold need to be determined. Then, the power spectrum characteristics are accumulated and averaged over N1 frames to eliminate fluctuations caused by noise, and the average value is used as the power spectrum characteristic parameter. The current squelch gate state is determined, and the power spectrum characteristic parameter is compared with the corresponding threshold value. Based on the comparison result, it is decided whether the current carrier squelch result remains unchanged or is flipped. Specifically, when the squelch gate state is open, the power spectrum characteristic parameter is compared with the carrier closing threshold. If the power spectrum characteristic parameter is less than the carrier closing threshold, the carrier squelch result is output as 0; if the power spectrum characteristic parameter is greater than the carrier closing threshold, the carrier squelch result is output as 1. When the squelch gate state is closed, the power spectrum characteristic parameter is compared with the carrier opening threshold. If the power spectrum characteristic parameter is greater than the carrier opening threshold, the carrier squelch result is output as 1; if the power spectrum characteristic parameter is less than the carrier opening threshold, the carrier squelch result is output as 0.

[0052] The voice squelch detection module first performs voice processing on the voice signal, including demodulation, DC removal, and filtering. Then, it extracts voice features from the processed voice signal, using the energy ratio of low and high frequency signals as the voice feature. Based on... Figure 5 The time-domain and frequency-domain plots of the shown voice signal reveal that the low-frequency energy is higher than the high-frequency energy, a characteristic that effectively distinguishes voice from noise. Therefore, the processed voice signal is framed, and each frame is input to two bandpass filters with identical gain. One bandpass filter outputs the high-frequency signal, and the other outputs the low-frequency signal. The squares of the moduli of the low-frequency and high-frequency signals are used as the time-domain energy value, and their ratio is calculated as the voice feature. If it is a voice signal, the low-frequency signal energy is higher than the high-frequency signal energy, so the ratio will be much greater than 1; if it is noise, the low-frequency and high-frequency signal energies are similar, and their ratio is close to 1. After accumulating N2 frames, the largest ratio is selected as the voice feature parameter for that segment of the signal.

[0053] The noise reduction decision process based on speech features is as follows: Figure 6As shown. First, the current squelch level needs to be locked, and the voice opening threshold and voice closing threshold need to be determined. The locked squelch level here is the same as the squelch level determined by the carrier squelch decision. Then, the maximum value of the voice features in frame N2 is calculated, and the maximum energy ratio is used as the voice feature parameter. The current squelch gate state is determined, and the voice feature parameter is compared with the corresponding threshold value. Based on the comparison result, it is decided whether the current voice squelch result remains unchanged or is flipped. Specifically, when the squelch gate state is open, the voice feature parameter is compared with the voice closing threshold. If the voice feature parameter is less than the voice closing threshold, the voice squelch result is output as 0; if the voice feature parameter is greater than the voice closing threshold, the voice squelch result is output as 1. When the squelch gate state is closed, the voice feature parameter is compared with the voice opening threshold. If the voice feature parameter is greater than the voice opening threshold, the voice squelch result is output as 1; if the voice feature parameter is less than the voice opening threshold, the voice squelch result is output as 0.

[0054] Combining carrier detection and voice detection technologies can reduce the probability of false alarms and missed alarms caused by single-feature judgment criteria. The carrier squelch results and voice squelch results calculated by the carrier detection and voice detection modules are jointly input into the joint decision module. Since the frame length of voice squelch detection is much shorter than that of carrier squelch detection, the frequency of voice squelch result reversal is greater than that of carrier detection results. This is because voice signals in the time domain frequently exhibit pauses or intermittent phenomena. At these pauses, voice features are indistinguishable from noise, leading to a judgment of zero for the voice squelch result. However, carrier squelch decision is based on the frequency domain, and brief pauses do not affect the power spectrum characteristics in the frequency domain. Therefore, during joint decision-making, it is necessary to observe for a period of time, such as approximately 500ms, to confirm that both the voice squelch result and the carrier squelch result are in a stable state within this observation period, thus avoiding frequent opening and closing of the squelch gate. The joint decision-making process is as follows: Figure 7 As shown. If both the carrier squelch result and the voice squelch result are 0 and remain stable during the observation period, the squelch gate can be closed; if it is unstable, the squelch gate state should be left unchanged. If both the carrier squelch result and the voice squelch result are 1, the squelch gate can be directly set to the open state; otherwise, the squelch gate state should be left unchanged.

[0055] It is understood that those skilled in the art can make equivalent substitutions or modifications to the technical solution and inventive concept of the present invention, and all such substitutions or modifications should fall within the protection scope of the appended claims.

Claims

1. A voice signal squelch device based on joint detection, comprising an AD module, a down-conversion module, an AGC module, a carrier squelch detection module, a voice squelch detection module, and a joint detection module, characterized in that: The received intermediate frequency voice signal first undergoes analog-to-digital conversion and digital down-conversion through the AD module and down-conversion module; then, after digital automatic gain control through the AGC module, the voice signal is divided into two paths; the carrier detection processing module directly performs carrier detection processing on one of the voice signals to obtain power spectrum characteristics; The voice squelch detection module performs AM / FM demodulation on another voice signal. On the one hand, it outputs the demodulated voice while the squelch gate is open. On the other hand, it performs voice feature detection processing on the demodulated voice to obtain voice features. The joint decision module performs a joint decision on the power spectrum features and voice features calculated by the carrier squelch detection module and the voice squelch detection module, and controls the opening and closing state of the squelch gate.

2. The voice signal noise reduction device based on joint detection according to claim 1, characterized in that... If the modulation mode of the carrier detection processing module is FM, the signal power within the signal bandwidth is selected as the main signal power; the signal power deviating from the signal bandwidth by about 50% is selected as the noise floor power. The ratio of the main signal power to the noise floor power is used as the power spectrum characteristic.

3. The voice signal squelch device based on joint detection according to claim 1, characterized in that... If the modulation mode of the carrier detection processing module is AM, the signal covered by the highest point of the power spectrum is taken as the main signal power, and the signal that deviates from the signal bandwidth by about 50% is selected as the noise floor power according to the same bandwidth range. The difference between the main signal power and the noise floor power is used as a power spectrum feature.

4. A voice signal noise reduction device based on joint detection according to claim 1, characterized in that... The voice squelch detection module first processes the voice signal, including demodulation, DC removal, and filtering. Then, it performs frame-by-frame processing on the processed voice signal and inputs it to two bandpass filters with the same gain. One bandpass filter outputs a high-frequency signal, and the other outputs a low-frequency signal. The square of the modulus of the low-frequency signal and the high-frequency signal is used as the time-domain energy value, and the ratio is processed to obtain the voice feature.

5. A voice signal noise reduction device based on joint detection according to claim 1, characterized in that... The joint decision-making process based on power spectrum features and speech features includes the following steps: Step 1: Squelch Decision Based on Power Spectrum Characteristics. First, the current squelch level is locked, and the carrier opening threshold and carrier closing threshold are determined. Then, the power spectrum characteristics are accumulated and averaged over N1 frames, and the average value is used as the power spectrum characteristic parameter. When the squelch gate is open, the power spectrum characteristic parameter is compared with the carrier closing threshold. If the power spectrum characteristic parameter is less than the carrier closing threshold, the carrier squelch result is output as 0; if the power spectrum characteristic parameter is greater than the carrier closing threshold, the carrier squelch result is output as 1. When the squelch gate is closed, the power spectrum characteristic parameter is compared with the carrier opening threshold. If the power spectrum characteristic parameter is greater than the carrier opening threshold, the carrier squelch result is output as 1; if the power spectrum characteristic parameter is less than the carrier opening threshold, the carrier squelch result is output as 0. Step 2: Squelch determination based on voice features. First, the current squelch level is locked, and the voice opening threshold and voice closing threshold are determined. Then, voice features of N2 frames are accumulated, and the largest voice feature is used as the voice feature parameter. When the squelch gate is open, the voice feature parameter is compared with the voice closing threshold. If the voice feature parameter is less than the voice closing threshold, the voice squelch result is output as 0; if the voice feature parameter is greater than the voice closing threshold, the voice squelch result is output as 1. When the squelch gate is closed, the voice feature parameter is compared with the voice opening threshold. If the voice feature parameter is greater than the voice opening threshold, the voice squelch result is output as 1; if the voice feature parameter is less than the voice opening threshold, the voice squelch result is output as 0. Step 3: Joint decision. If both the carrier squelch result and the voice squelch result are 0 and remain stable during the observation period, the squelch gate is closed. If it is unstable, the squelch gate is left as is. If both the carrier squelch result and the voice squelch result are 1, the squelch gate can be directly set to the open state. In other cases, the squelch gate is left as is.

Images