CN105244040B - An audio signal comparative process consistency - Google Patents

An audio signal comparative process consistency Download PDF

Info

Publication number
CN105244040B
CN105244040B CN201510429221.5A CN201510429221A CN105244040B CN 105244040 B CN105244040 B CN 105244040B CN 201510429221 A CN201510429221 A CN 201510429221A CN 105244040 B CN105244040 B CN 105244040B
Authority
CN
China
Prior art keywords
data
similarity
audio signal
path
consistency
Prior art date
Application number
CN201510429221.5A
Other languages
Chinese (zh)
Other versions
CN105244040A (en
Inventor
王世为
吕连新
赵凡
Original Assignee
杭州联汇数字科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 杭州联汇数字科技有限公司 filed Critical 杭州联汇数字科技有限公司
Priority to CN201510429221.5A priority Critical patent/CN105244040B/en
Publication of CN105244040A publication Critical patent/CN105244040A/en
Application granted granted Critical
Publication of CN105244040B publication Critical patent/CN105244040B/en

Links

Abstract

本发明公开了种音频信号致性对比方法,其包括以下步骤:获取采样数据,扩展长度,计算快速相关相似性,计算相似度及偏移量,如果相似度大于或等于阈值则判定两路数据致,如果相似度小于阈值则判断采样数据是否已经进行近似包络线处理,如果已经进行近似包络线处理则判定两路数据不致,如果未进行近似包络线处理则对采样数据进行近似包络线处理后重复前述步骤。 The present invention discloses a method of comparative activated species-audio signal, comprising the steps of: obtaining the sampled data, the extended length, fast correlation calculating the similarity, and the similarity is calculated offset, if the similarity is greater than or equal to the threshold value it is determined that two-way data induced, if the similarity is smaller than the threshold is determined whether the sampled data have been approximated envelope processing, if already approximated envelope processing is determined that the two-way data will not, if the approximate envelope without processing the sampled data is approximated packet repeating the foregoing steps after envelope processing. 本方案可以准确判断两路音频信号的致性,适用于广播电台对播出信号的监控。 The present embodiment can accurately determine the consistency of two audio signals, suitable monitoring of the radio broadcast signal.

Description

一种音频信号一致性对比方法 An audio signal comparative process consistency

技术领域 FIELD

[0001] 本发明涉及音频数据分析处理领域,尤其是涉及一种音频信号一致性对比方法。 [0001] The present invention relates to the field of analysis processing audio data, more particularly to a method for audio signals consistent comparisons. 背景技术 Background technique

[0002] 在广播电台的整个传输、监控链路中,都是针对音频的物理指标进行检测的,比如:是否停播、电平偏低或偏高、反相等等。 [0002] Throughout transmission of broadcast stations, monitoring the link, is for detecting a physical audio indicators, such as: whether or off the air, low or high level, an inverter and the like. 但对于音频的内容是否正确缺少检测手段,例如总控最后输出的信号是否是直播室调音台输出的信号,空中发射信号是否是总控末级输出的信号等问题目前没有合适的检测手段,只有等发生了事故才能查知。 But for the audio content is correct the lack of means of detection, such as master control signal whether the final output of the mixer is the broadcast room output signal, the transmitted signal is a signal whether the air master control issues such as the final output is currently no suitable means of detection, only wait for an accident to check on.

[0003] 中华人民共和国国家知识产权局于2010年12月01日公开了公开号为CN101902677A的专利文献,名称是音频检测装置及方法,该装置包括:MIC电路检测模块和发声电路检测模块;MIC电路检测模块,用于获取MIC电路采样后的正弦波信号的幅度与频率,根据幅度与频率确定MIC电路是否合格,并输出正弦波信号;发声电路检测模块包括:具有MIC电路的转换板以及处理器;转换板,用于采样发声电路输出的模拟信号,并将模拟信号转换为数字信号,发送至处理器;处理器,用于对数字信号的幅度与频率进行分析处理,判断发声电路是否合格。 [0003] State Intellectual Property Office on December 1, 2010 discloses Publication No. CN101902677A patent literature, the name is an audio detection device and method, the apparatus comprising: MIC circuit detection module and a detection circuit sound module; MIC circuit detecting module, configured to obtain amplitude and frequency of the sine wave signal sampling circuit MIC, MIC is determined according to the amplitude and frequency of the circuit is qualified, and outputs a sine wave signal; utterance detection circuit module comprising: a plate having a conversion processing circuit and MIC ; a converter board for sampling the analog signal outputted from the sound circuit, the analog signal into a digital signal, transmitted to the processor; processor for amplitude and frequency analysis of the digital signal processing circuit determines whether the utterance eligibility . 此方案无法对音频信号进行一致性的判断。 This program can not be judged on the consistency of the audio signal.

发明内容 SUMMARY

[0004] 本发明主要是解决现有技术所存在的不能对音频信号一致性进行准确判断的技术问题,提供一种可以通过对音频的特征信号分析,检测两路信号内容是否一致的音频信号一致性对比方法,便于广播电台及时发现播出故障。 [0004] The present invention is mainly to solve the prior art problems existing technology can not accurately determine the consistency of the audio signal, a method can be characterized by analysis of the audio signal, detecting whether the two signals consistent with the content of the same audio signal comparison of methods to facilitate timely detection of radio stations broadcast fault.

[0005] 本发明针对上述技术问题主要是通过下述技术方案得以解决的:一种音频信号一致性对比方法,包括以下步骤: [0005] The present invention addresses the above technical problem is solved by the following technical scheme: An audio signal comparison consistency method, comprising the steps of:

[0006] S01、从第一路音频信号中获取N个采样数据作为第一路采样数据,从第二路音频信号中获取N个采样数据作为第二路采样数据; [0006] S01, acquires N samples of data from the first audio channel signal as a first sampled data, the second audio signal obtained from the N sample data sample as the second data path;

[0007] S02、将第一路采样数据后面补0扩展至长度为2N,将第二路采样数据后面补0扩展至长度为2N; [0007] S02, the sampling data following the first path extended up to a length of 0 2N, the second sample data paths back to 0s extended length 2N;

[0008] S03、对扩展后的第一路采样数据进行快速傅里叶变换然后求共辄得到第一路数据,对扩展后的第二路采样数据进行快速傅里叶变换然后求共辄得到第二路数据,求第一路数据和第二路数据的互相关函数,在对互相关函数进行快速傅里叶变换,得到相似性数据; [0008] S03, the first sample-and-extended data and then find the fast Fourier transform to obtain a first common path data Noir, a second sample-and-extended data and then find the fast Fourier transform to obtain a total Noir a second data path, the first path data and second request data path correlation function, the cross-correlation function of the fast Fourier transform, to obtain similarity data;

[0009] S04、选取相似性数据中绝对值的最大值,然后计算相似度值和最大值偏移量; [0009] S04, select the maximum absolute value of the similarity data, and then calculating the similarity value and the maximum offset;

[0010] S05、如果相似度值大于或等于一致性阈值,则认为第一路采样数据和第二路采样数据一致,对比流程结束;如果相似度小于一致性阈值,则进入步骤S06; [0010] S05, if the similarity value is greater than or equal to the threshold consistency, is considered consistent with a first passage and a second sample data sampled data path, comparison process ends; if the similarity is smaller than the threshold value the consistency, the process proceeds to step S06;

[0011] S06、判断第一路采样数据和第二路采样数据是否已经经过近似包络线处理,如果没有,则对第一路采样数据和第二路采样数据进行近似包络线处理,然后重复步骤S02至S05;如果已经经过近似包络线处理,则判定第一路采样数据和第二路采样数据不一致。 [0011] S06, determines whether the first data channel and the second channel sample data sampling has elapsed approximate envelope processing, and if not, the first passage and the second passage sampled data sampled data approximate envelope processing, and repeat steps S02 to S05; if the envelope has been processed through similar package, it is determined that the first path inconsistent sampled data and the second channel sample data.

[0012]上述过程可以较为准确地判断第一路采样信号和第二路采样信号的相似度。 [0012] The process can be more accurately determined similarity first sampling signal and the second branch passage of the sampled signal. 根据计算出的偏移量对音频信号偏移后,可以更为准确地计算音频信号的一致性。 According to the calculated offset amount offset for the audio signal, the audio signal can be calculated more accurately consistent manner.

[0013]作为优选,相似度值ratio计算公式如下: [0013] Preferably, the similarity value ratio calculated as follows:

[0014] ratio = fSamplemax*M/sum [0014] ratio = fSamplemax * M / sum

[0015]式中,fSample·为相似性数据中绝对值的最大值,M为相似性数据的个数,sum为所有相似性数据的绝对值的和。 [0015] In the formula, fSample · is the maximum value of the absolute value of the similarity data, M being the number of similarity data, sum of absolute values ​​of all similarity data.

[0016]作为优选,最大值偏移量off set根据以下方法确定:f Samplemax为相似性数据中的第1个数据,如果1小于或等于1/2,则0打86〖=;[;如果;[大于1/2,则0打86〖=;[-]\1/2;f Samp I emx为相似性数据中绝对值的最大值,M为相似性数据的个数。 [0016] Advantageously, the maximum value of the offset is determined according to the following method off set: f Samplemax similarity data is first data, if less than 1 or equal to 1/2, then playing 86 〖= 0; [; if ; [greater than 1/2, the play 86 〖= 0; [-] \ 1/2; f Samp I emx maximum similarity data for the absolute value, M being the number of similarity data.

[0017] 作为优选,近似包络线处理具体为:将第一路采样数据划分为长度为2毫秒的窗口,从每个窗口中取最大值形成第一窗口最大数据*X1;将第二路采样数据划分为长度为2毫秒的窗口,从每个窗口中取最大值形成第二窗口最大数据*Y1;计算第一路采样数据的绝对值的均值meanXl;计算第二路采样数据的绝对值的均值meanYl;将*X1中的各个数据减去meanXl,所获得的数据作为第一路采样数据;将*Y1中的各个数据减去meanYl,所获得的数据作为第二路采样数据。 [0017] Advantageously, the approximate envelope processing specifically comprises: a first sampled data into a path length of 2 ms window, each window taking a maximum value is formed from a first maximum window data * X1; a second passage sampled data into a length of 2 ms window, each window taking a maximum value is formed from the second window maximum data * Y1; calculating a first sampled data channel mean absolute meanXl; second absolute value calculation sampled data path mean meanYl; * the respective data X1 is subtracted meanXl, the obtained data as the first data sample path; * Yl respective data subtracted meanYl, the obtained data as the second channel sample data.

[0018] 对于音量比较小且底噪明显的音频内容,直接进行步骤S02-S05的计算得出的相似度比较小,一般会小于一致性阈值,此时进行近似包络线处理后可以消除小信号干扰,提高判别率。 After [0018] For relatively small volume and a significant noise floor audio content directly calculation step S02-S05 derived similarity is relatively small, usually less than the consistency threshold, this time is approximated envelope could be eliminated by small signal interference, increasing the recognition rate.

[0019] 作为优选,所述一致性阈值的取值范围为20-30。 [0019] Advantageously, the consistency of the threshold ranges from 20-30.

[0020] 通过一定量的采样分析后可以定位出准确的一致性阈值。 [0020] consistency can accurately locate the threshold by the amount of sampling and analysis.

[0021] 作为优选,当连续三次判定第一路采样数据和第二路采样数据不一致则认定第一路音频信号和第二路音频信号不一致。 [0021] Advantageously, the first channel is determined when three consecutive sampling data and the second channel sample data inconsistency is asserted a first audio signal and the second audio signal inconsistencies.

[0022] 当出现第一路音频信号和第二路音频信号不一致的情况时,可以进行预警,提示工作人员进行修正或维护。 [0022] When the first audio signal and the second audio signal path inconsistent situation, can be early warning, prompt correction or maintenance staff.

[0023] 本方案可以用于以下场合: [0023] The present embodiment may be used in the following situations:

[0024] 电台总控: [0024] Radio Master Control:

[0025] 用于判断总控末级输出信号和调音台输出信号的一致性; [0025] The master control for determining the consistency of the final output signal and a mixer output signal;

[0026] 总控末级信号和空收信号的一致性; [0026] The total final consistency control signal and the empty signal is received;

[0027] 发射台: [0027] The transmitter:

[0028] 空收信号和光端机输出信号的一致性; [0028] Optical coherence null received signal and the output signal;

[0029] 监测中心: [0029] Monitoring Center:

[0030] 用户判断播出信号是否被干扰或者侵入; [0030] The user determines whether the broadcast signal is disturbed or invasion;

[0031] 本发明带来的实质性效果是,可以可靠地计算出两路音频信号的相似度以及偏移量,消除了音频信号衰减的影响,具有极高的判断准确度。 [0031] The invention offers substantial effect, it can be reliably calculated similarity and offsets two audio signals, the influence of the attenuation of the audio signal, is determined with high accuracy.

附图说明 BRIEF DESCRIPTION

[0032] 图1是本发明的一种流程图。 [0032] FIG. 1 is a flowchart illustrating the present invention.

具体实施方式 Detailed ways

[0033] 下面通过实施例,并结合附图,对本发明的技术方案作进一步具体的说明。 [0033] The following Examples, in conjunction with the accompanying drawings, technical solutions of the present invention will be further specifically described.

[0034] 实施例:本实施例的一种音频信号一致性对比方法,如图1所示,包括以下步骤: [0034] Example: Comparative consistent method of an audio signal according to the present embodiment, shown in Figure 1, comprising the steps of:

[0035] 录制一定采样点个数的音频数据,首先扩展长度为两倍,将后面的赋值为0。 [0035] recording a certain number of points sampled audio data, twice the first extension length, the second assignment is 0.

[0036] float fSample[N*2] [0036] float fSample [N * 2]

[0037] fSample [i] =0 N<i<N*2 [0037] fSample [i] = 0 N <i <N * 2

[0038] 然后将采样数据进行FFT变换,FFT算法描述如下: [0038] The sampled data is then FFT transform, FFT algorithm is described as follows:

[0039] 根据离散傅氏变换的奇、偶、虚、实等特性,对离散傅立叶变换的算法进行改进获得快速傅氏变换算法FFT,设X (η)为N项的复数序列,由DFT变换,任一X (m)的计算都需要N次复数乘法和N-1次复数加法,而一次复数乘法等于四次实数乘法和两次实数加法,一次复数加法等于两次实数加法,即使把一次复数乘法和一次复数加法定义成一次“运算”(四次实数乘法和四次实数加法),那么求出N项复数序列的X (m),S卩N点DFT变换大约就需要N~2次运算。 [0039] The odd, even, virtual, real characteristics such discrete Fourier transform algorithm discrete Fourier transform is performed to improve access to the fast Fourier transform algorithm an FFT, provided X (η) is a complex sequence of N entries, the DFT of calculating any of X (m) are required N complex multiplications and N-1 complex additions, and a complex multiplication is equal to four real multiplications and two real additions, a complex addition is equal to two real additions, even if the time a complex multiplication and complex addition is defined as an "operation" (four real multiplications and real additions four times), then obtains complex sequence X (m) N terms, S N-point DFT of approximately Jie requires N ~ 2 times operations. 在FFT中,利用WN的周期性和对称性,把一个N项序列(设N = 2k,k为正整数),分为两个N/2项的子序列,每个N/2点DFT变换需要(N/2) 2次运算,再用N次运算把两个N/2点的DFT变换组合成一个N点的DFT变换。 In the FFT, using the periodicity and symmetry of WN, the N items of a sequence (set N = 2k, k a positive integer) is divided into two sub-sequences of N / 2 terms, each of N / 2-point DFT transform It requires (N / 2) times 2 operation, and then the combination of N times the DFT arithmetic two N / 2-point is transformed into an N-point DFT. 这样变换以后,总的运算次数就变成Ν+2* (Ν/2Γ2 = Ν+ (N~2) /2,节省了大约50%的运算量。 After this conversion, the total number of operations becomes Ν + 2 * (Ν / 2Γ2 = Ν + (N ~ 2) / 2, about 50% of savings in computation.

Figure CN105244040BD00051

[0043] Ι^η^ΞΝ [0043] Ι ^ η ^ ΞΝ

[0044] 将FFT预算后的两路采样数据,求共辄,再进行一次FFT运算,过程称为计算快速相关相似性。 [0044] The two sampled data after FFT budget, a total seek Noir, then a FFT operation, a process called rapid calculation of correlation similarity.

[0045] 然后将傅氏变换的数据取最大值,通过采样个数计算最大平均值,并通过最大值的开始位计算最大值偏移量,这样就可以计算出相似度值及两路音频的采样点偏移量,计算方法为: [0045] and the Fourier transform of the data takes the maximum value, is calculated by averaging the maximum number of samples, and the maximum value of the start position shift amount calculated by the maximum value, so that values ​​can be calculated and the similarity of the two audio sampling point shift amount calculated as:

[0046] 采样点个数:N [0046] The number of sampling points: N

[0047] 偏移量:offset [0047] offset: offset

[0048] 相似度:ratio [0048] Similarity: ratio

[0049] 傅氏变换后数据:fSample,因为音频信号为正弦波,有正负值,所以计算的时候应当使用绝对值进行大小对比。 [0049] Fourier transformed data: fSample, since the audio signal is a sine wave with positive and negative values, so should be used when the absolute value of the calculated size comparison.

[0050] 数据量总和:sum The sum of [0050] data amount: sum

[0051] ratio< I fSample [i] I ?ratio= I fSample [i] I offset = i [0051] ratio <I fSample [i] I? Ratio = I fSample [i] I offset = i

[0052] Sum = Sum+I fSample [i] [0052] Sum = Sum + I fSample [i]

[0053] l^i^N [0053] l ^ i ^ N

[0054] [0054]

Figure CN105244040BD00052

[0055] [0055]

Figure CN105244040BD00061

[0056] 当相似度(ratio)值大于一定的范围值(经过长期的测试后该值一般定义为20-30)后,就可以认定为两路信号一致,然后再根据计算出的偏移量(offset),进行两路信号的偏移,偏移算法为: After [0056] When the degree of similarity (ratio) is greater than a certain range of values ​​(after long-term test of the value is generally defined as 20-30), can be identified as the two signals are coincident, and then based on the calculated shift amount (offset), two signals offset, offset algorithm:

[0057] fSample [i] =fSample [i+offset] [0057] fSample [i] = fSample [i + offset]

[0058] (N-〇ffset) [0058] (N-〇ffset)

[0059] 后续对偏移过的数据进行计算就会相似度值更加贴近对比范围,提高准确度。 [0059] subsequent to the offset data will be calculated through the similarity value closer comparison range, improve accuracy.

[0060] 对于音量比较小,且底噪明显的音频内容,在当前算法得出的近似值比较小,一般会小于预定的范围,该值波动情况明显,且相似度值普遍在10-15左右,该情况时,采用取音频线路的近似包络线的方式进行二次计算,可以消除小信号干扰提高判别率。 [0060] For relatively small volume, and the noise floor of the audio content significantly, resulting in an approximation of the current algorithm is relatively small, usually less than a predetermined range, the value is significant fluctuations, and generally in the 10-15 similarity values, in this case, by way of the approximate envelope taken for secondary audio line calculations, the interference signal can be eliminated to improve the recognition rate and small. 每次取2毫秒窗口的最大值,得到*xi、*Y1,及两路信号的均值: Takes the maximum value of 2 milliseconds per window, to give * xi, * mean Y1, and two signals:

[0061] 设定采样率为48000,那么2毫秒的窗口采样数为96 [0061] 48 000 to set the sampling rate, then the window sample number 2 to 96 milliseconds

[0062] [0062]

Figure CN105244040BD00062

[0063] Ι^Ξΐ^ΞΝ; i+ = 96; j++ [0063] Ι ^ Ξΐ ^ ΞΝ; i + = 96; j ++

[0064] sum = sum+I f SampIe [i] [0064] sum = sum + I f SampIe [i]

[0065] l^i^N [0065] l ^ i ^ N

[0066] [0066]

Figure CN105244040BD00063

[0067] 将*X1、*Y1信号中的均值去掉: [0067] The mean * X1, * Y1 signal is removed:

[0068] rX = *Xl_meanXl [0068] rX = * Xl_meanXl

[0069] rY = *Yl_meanYl [0069] rY = * Yl_meanYl

[0070] 然后对二次计算的数据rX、rY重新之前的步骤,计算快速相关相似性,对计算出的数据再次进行相似度及偏移量的计算,获取新的对比结果,进一步提高准确度。 [0070] The data is then calculated secondary rX, rY re previous step, fast correlation calculating the similarity, the calculated similarity is calculated and the data offset again, obtaining new comparison results, to further improve the accuracy . 由于这类音频在实际环境中出现的频率并不高,而且两次计算会增进CPU负担,故是作为相似度值较低的时候的二次验证策略存在。 Since the frequency of occurrence of this type of audio in a real environment is not high, but the two calculations will enhance CPU burden, it is as low similarity value when the secondary verification policy exists.

[0071] 在快速相关相似性的基础上再加入判断策略,连续2次判断阀值小于设定值(一般为20-30),且偏移量波动范围较大(大于Is的采样)时即认为两路信号不一致。 When [0071] added on the basis of similarity of the fast correlation judgment policy, 2 times less than the set determination threshold value (typically 20-30), and the offset fluctuation range (greater than Is sampling) i.e. found to be inconsistent two signals.

[0072] 经过快速相关相似性计算之后,可以可靠的计算出两路音频的相似度及偏移量,再通过一定量的采样分析后,就可以定位出准确的判断阀值,根据此阀值便可以有效的对两路信号的一致性进行检测、预警,并且由于进行了两次FFT运算,取相似度是进行的信号曲线的计算,音频信号的衰减并不影响计算结果,判断准确度可以达到99.99%以上。 [0072] After fast correlation similarity calculation, can be reliably calculated and the similarity of two audio offset, and then after a certain amount through the analysis, can locate the exact determination threshold, based on this threshold value consistency can effectively detect two signals, warning, and as a result of the FFT operation twice, taking the similarity is calculated, the attenuation of the audio signal is a signal does not affect the calculation of the curve, the accuracy of the determination can be 99.99%.

[0073] 本文中所描述的具体实施例仅仅是对本发明精神作举例说明。 Specific Example [0073] described herein is merely illustrative for spirit of the invention. 本发明所属技术领域的技术人员可以对所描述的具体实施例做各种各样的修改或补充或采用类似的方式替代,但并不会偏离本发明的精神或者超越所附权利要求书所定义的范围。 Those skilled in the art of the present invention can be made to the specific embodiments described various modifications or additions, or a similar alternative embodiment, but without departing from the spirit of the invention or exceed defined in the appended claims range.

[0074] 尽管本文较多地使用了采样数据、快速傅里叶变换等术语,但并不排除使用其它术语的可能性。 [0074] While the terms are used herein, sample data, etc. The term fast Fourier transform, but do not exclude the possibility of using other terminology. 使用这些术语仅仅是为了更方便地描述和解释本发明的本质;把它们解释成任何一种附加的限制都是与本发明精神相违背的。 These terms are only used to make it easier to describe and explain the essence of the present invention; interprets them into any additional restrictions are contrary to the spirit of the invention.

Claims (6)

1. 一种音频信号一致性对比方法,其特征在于,包括以下步骤: 501、 从第一路音频信号中获取N个采样数据作为第一路采样数据,从第二路音频信号中获取N个采样数据作为第二路采样数据; 502、 将第一路采样数据后面补O扩展至长度为2N,将第二路采样数据后面补O扩展至长度为2N; 503、 对扩展后的第一路采样数据进行快速傅里叶变换然后求共辄得到第一路数据,对扩展后的第二路采样数据进行快速傅里叶变换然后求共辄得到第二路数据,求第一路数据和第二路数据的互相关函数,在对互相关函数进行快速傅里叶变换,得到相似性数据; 504、 选取相似性数据中绝对值的最大值,然后计算相似度值和最大值偏移量; 505、 如果相似度值大于或等于一致性阈值,则认为第一路采样数据和第二路采样数据一致,对比流程结束;如果相似度小于一致性阈值, An audio comparison method consistency, characterized by comprising the steps of: 501, a first audio signal obtained from the N sample data sampled data as a first passage, a second passage from the N acquired audio signal a second sample data sampled data channel; 502, sampled data following the first path extends to a length of up O 2N, the extension of the second channel sample data back up to a length of O 2N; 503, a first passage of the extended sampled data and request a fast Fourier transform to obtain a first common path data Noir, a second sample-and-extended data and then find the fast Fourier transform to obtain a second common path data Noir, seeking a first data path and Road data of the cross-correlation function, the cross-correlation function of the fast Fourier transform, to obtain similarity data; 504, selecting the maximum absolute value of the similarity data, and then calculating the similarity value and the maximum offset; 505, if the similarity value is greater than or equal to the threshold consistency, is considered consistent with a first passage and a second sample data sampled data path, comparison process ends; if the similarity is smaller than the threshold value consistency, 进入步骤S06; 506、 判断第一路采样数据和第二路采样数据是否已经经过近似包络线处理,如果没有,则对第一路采样数据和第二路采样数据进行近似包络线处理,然后重复步骤S02至S05; 如果已经经过近似包络线处理,则判定第一路采样数据和第二路采样数据不一致。 Proceeds to step S06; 506, determines whether the first channel sample data and the second channel sample data has elapsed approximate envelope processing, and if not, the first passage and the second sample data sampled data path envelope approximation process, then repeat steps S02 to S05; if the envelope has been processed through similar package, it is determined that the first path inconsistent sampled data and the second channel sample data.
2. 根据权利要求1所述的一种音频信号一致性对比方法,其特征在于,相似度值ratio 计算公式如下: rat io = f Sampl emax*M/ sum 式中,f Samplemax为相似性数据中绝对值的最大值,M为相似性数据的个数,sum为所有相似性数据的绝对值的和。 2. An audio signal consistent comparisons method according to claim 1, characterized in that the ratio a similarity value is calculated as follows: rat io = f Sampl emax * M / sum formula, f Samplemax as similarity data maximum absolute value, M being the number of similarity data, sum of absolute values ​​of all similarity data.
3. 根据权利要求1所述的一种音频信号一致性对比方法,其特征在于,最大值偏移量off set根据以下方法确定:f Samplemax为相似性数据中的第i个数据,如果i小于或等于M/2, 则off set = i ;如果i大于M/2,则off set = iM/2;f Samplemax为相似性数据中绝对值的最大值,M为相似性数据的个数。 An audio signal according to the consistency of the comparative method in claim 1, characterized in that the maximum value of the offset is determined according to the following method off set: f Samplemax as similarity data in the i-th data, if i is less than or equal to M / 2, then the off set = i; if i is larger than M / 2, then the off set = iM / 2; f Samplemax maximum absolute value of the similarity data, M being the number of similarity data.
4. 根据权利要求2或3所述的一种音频信号一致性对比方法,其特征在于,近似包络线处理具体为:将第一路采样数据划分为长度为2毫秒的窗口,从每个窗口中取最大值形成第一窗口最大数据*X1;将第二路米样数据划分为长度为2毫秒的窗口,从每个窗口中取最大值形成第二窗口最大数据*Y1;计算第一路采样数据的绝对值的均值meanXl;计算第二路采样数据的绝对值的均值meanYl;将*X1中的各个数据减去meann,所获得的数据作为第一路采样数据;将*Y1中的各个数据减去meanYl,所获得的数据作为第二路采样数据。 4. An audio signal consistent comparisons 2 or claim 3, characterized in that the envelope approximation process specifically includes: a first sampled data into a path length of 2 ms window, from each of forming a first window, taking the maximum window maximum data * X1; m second path length of the sample data is divided into 2 ms window, each window taking a maximum value formed from the second window maximum data * Y1; calculating a first mean absolute meanXl path sample data; calculating a second channel sample data mean absolute meanYl; * the respective data X1 is subtracted meann, the obtained data as the first data sample path; Yl of the * subtracting the individual data meanYl, the obtained data as the second channel sample data.
5. 根据权利要求1所述的一种音频信号一致性对比方法,其特征在于,所述一致性阈值的取值范围为20-30。 A comparative method of an audio signal according to the consistency of claim 1, wherein said consistency threshold ranges 20-30.
6. 根据权利要求1所述的一种音频信号一致性对比方法,其特征在于,当连续三次判定第一路采样数据和第二路采样数据不一致则认定第一路音频信号和第二路音频信号不一致。 An audio signal according to the consistency of the comparative process of claim 1, wherein, when three successive sampling data determined first branch passage and the second sample data inconsistent with the first audio signal is asserted and the second channel audio inconsistent signals.
CN201510429221.5A 2015-07-20 2015-07-20 An audio signal comparative process consistency CN105244040B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201510429221.5A CN105244040B (en) 2015-07-20 2015-07-20 An audio signal comparative process consistency

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201510429221.5A CN105244040B (en) 2015-07-20 2015-07-20 An audio signal comparative process consistency

Publications (2)

Publication Number Publication Date
CN105244040A CN105244040A (en) 2016-01-13
CN105244040B true CN105244040B (en) 2018-12-11

Family

ID=55041659

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201510429221.5A CN105244040B (en) 2015-07-20 2015-07-20 An audio signal comparative process consistency

Country Status (1)

Country Link
CN (1) CN105244040B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108259104A (en) * 2017-12-08 2018-07-06 中国航空工业集团公司成都飞机设计研究所 Dual-redundancy transmitting radio station voting method

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US3919479A (en) * 1972-09-21 1975-11-11 First National Bank Of Boston Broadcast signal identification system
CN102142257A (en) * 2010-12-28 2011-08-03 北京北大方正电子有限公司 Audio signal processing method and device
CN103426439A (en) * 2013-05-08 2013-12-04 杭州联汇数字科技有限公司 Method for detecting consistency of broadcast television audio signal content
CN104468697A (en) * 2014-10-10 2015-03-25 浙江广播电视集团 Radio station data transmission load distribution method
CN104505101A (en) * 2014-12-24 2015-04-08 北京巴越赤石科技有限公司 Real-time audio comparison method

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US3919479A (en) * 1972-09-21 1975-11-11 First National Bank Of Boston Broadcast signal identification system
CN102142257A (en) * 2010-12-28 2011-08-03 北京北大方正电子有限公司 Audio signal processing method and device
CN103426439A (en) * 2013-05-08 2013-12-04 杭州联汇数字科技有限公司 Method for detecting consistency of broadcast television audio signal content
CN104468697A (en) * 2014-10-10 2015-03-25 浙江广播电视集团 Radio station data transmission load distribution method
CN104505101A (en) * 2014-12-24 2015-04-08 北京巴越赤石科技有限公司 Real-time audio comparison method

Also Published As

Publication number Publication date
CN105244040A (en) 2016-01-13

Similar Documents

Publication Publication Date Title
CN1622200B (en) Method and apparatus for multi-sensory speech enhancement
CN1110034C (en) Spectral subtraction noise suppression method
JP4219774B2 (en) Nonlinear observation model to remove noise from the degraded signal
CN101236250B (en) Sound determination method and sound determination apparatus
CN101010722B (en) Device and method of detection of voice activity in an audio signal
JP4757158B2 (en) Sound signal processing method, the sound signal processing apparatus and a computer program
KR100883712B1 (en) Method of estimating sound arrival direction, and sound arrival direction estimating apparatus
EP2633524B1 (en) Method, apparatus and machine-readable storage medium for decomposing a multichannel audio signal
JP3812887B2 (en) Signal processing system and method
US8213263B2 (en) Apparatus and method of detecting target sound
JP3588030B2 (en) Speech segment determination apparatus and speech segment determination method
CN102708874A (en) Noise adaptive beamforming for microphone arrays
US6349278B1 (en) Soft decision signal estimation
EP1774517B1 (en) Audio signal dereverberation
CN101339769B (en) Echo suppressor and echo suppressing method
JP5000647B2 (en) Multi-sensor sound high quality using the voice state model
CN104252469B (en) A method for pattern matching, devices and circuits
JP4587160B2 (en) Signal processing apparatus and method
US8775173B2 (en) Erroneous detection determination device, erroneous detection determination method, and storage medium storing erroneous detection determination program
US9084049B2 (en) Automatic equalization using adaptive frequency-domain filtering and dynamic fast convolution
KR101562904B1 (en) Each arrival measurement apparatus and method
KR101422844B1 (en) Multi-sensory speech enhancement using a clean speech prior
CN102938649B (en) Adaptive Reconstruction decompression method based on compression power quality perceived Theory
CN101674639A (en) Detection method, device and system of channel quality
CN104132250B (en) Leakage one feature vector extraction duct improved wavelet packet based

Legal Events

Date Code Title Description
C06 Publication
C10 Entry into substantive examination
GR01 Patent grant
CP03 Change of name, title or address