阅读说明：本技术 一种基于半监督学习的声信号非视距识别方法 (A kind of acoustical signal non line of sight recognition methods based on semi-supervised learning ) 是由 胡志新 张磊 白旭晶 钟宇 薛文涛 左文斌 焦侃 杨伟婷 王楠 于 2019-08-20 设计创作，主要内容包括：本发明公开了一种基于半监督学习的声信号非视距识别方法,采集原始声信号x[n],对采集的原始声信号x[n]进行探测及分割,获得互相关结果片段R<Sub>i</Sub>[τ],对得到的互相关结果片段R<Sub>i</Sub>[τ]进行特征提取及非视距识别。能够获取有标签声信号数据样本和无标签数据样本,并提取出声信号数据样本的多个特征,然后基于这些特征值利用半监督学习进行非视距识别。本发明方法根据少量已知类别的声信号数据,自动区分大量未知声信号数据,本发明方法不必获取大量训练数据,节省了人力物力,且分类识别效果较好,解决了只有少量已知样本情况下声信号非视距识别的难题,为基于声技术的室内定位系统的实际应用提供了基础。(The acoustical signal non line of sight recognition methods based on semi-supervised learning that the invention discloses a kind of acquires original sound signals x [n], the original sound signals x [n] of acquisition is detected and divided, and obtains cross correlation results segment R i [τ], to obtained cross correlation results segment R i [τ] carries out feature extraction and non line of sight identification.Label acoustical signal data sample can be obtained and without label data sample, and extract multiple features of acoustical signal data sample, be then based on these characteristic values using semi-supervised learning and carry out non line of sight identification.The method of the present invention is according to the acoustical signal data of a small amount of known class, the a large amount of unknown acoustical signal data of automatic distinguishing, the method of the present invention need not obtain a large amount of training datas, save manpower and material resources, and Classification and Identification effect is preferable, acoustical signal non line of sight identifies in the case of solving the problems, such as only a small amount of known sample, and the practical application for the indoor locating system based on audio technology provides the foundation.)

1. a kind of acoustical signal non line of sight recognition methods based on semi-supervised learning, which comprises the following steps:

S1: acquisition original sound signals x [n]；

S2: being detected and divided to the original sound signals x [n] acquired in S1, and cross correlation results segment R is obtained_i[τ]；

S3: to cross correlation results segment R obtained in S2_i[τ] carries out feature extraction and non line of sight identification.

2. a kind of acoustical signal non line of sight recognition methods based on semi-supervised learning according to claim 1, which is characterized in that S2 includes the following steps:

S2.1: original sound signals x [n] is filtered and is enhanced, enhanced acoustical signal x'[n is obtained]；

S2.2: construction reference signal r [n], using reference signal r [n] to enhanced acoustical signal x'[n] carry out cross-correlation meter It calculates, obtains result R_x'r[τ]；

S2.3: to the result R obtained in S2.2_x'r[τ] is detected and is split extraction, obtains cross correlation results segment R_i [τ], the cross correlation results segment of i-th of enhanced acoustical signal of note are R_i[τ]。

3. a kind of acoustical signal non line of sight recognition methods based on semi-supervised learning according to claim 2, which is characterized in that

In S2.1, x'[n]=IFFT { FFT { x [n] } w [n] }, wherein w [n] is window function；

In S2.2,Wherein N be x'[n] length；

The specific method is as follows by S2.3:

To R_x'r[τ] carries out sequential detection, sets the length of sequential Load Signal segment as T_s, sequential Load Signal segment is seg [τ]=R_x'r[τ_s], wherein τ_s=[(i-1) T_s+1:iT_s]；Decision procedure in seg [τ] comprising useful signal is K { seg [τ] } >=thd, wherein thd is decision threshold, and K { } is that waveform kurtosis calculates symbol；If in seg [τ] including useful signal, according to letter The broadcasting timeline of mark node matches sequential Load Signal segment and cross correlation results segment with the ID of beaconing nodes, as a result It is denoted as a_i；The peak-peak position in cross correlation results segment is calculated, is denoted asIntercept acoustical signal and The subscript of cross correlation results segment indexes are as follows:

Beaconing nodes a_iThe signal segment x ' of acoustical signal_i[n]=x'[idx_s:idx_e], cross correlation results segment R_i[τ]=R_x'r [idx_s:idx_e]。

4. a kind of acoustical signal non line of sight recognition methods based on semi-supervised learning according to claim 2, which is characterized in that The window function is the compound window function of rectangular window and Blackman window composition, utilizes the length of rectangular windowTo carry out bandpass filtering to original sound signals x [n].

5. a kind of acoustical signal non line of sight recognition methods based on semi-supervised learning according to claim 1, which is characterized in that

S3 includes the following steps:

S3.1: to cross correlation results segment R_iRelative gain-the time delay distribution of [τ] is estimated, { Γ is obtained_a,Γ_τ}；

S3.2: { the Γ obtained from S3.1_a,Γ_τIn extract the characteristic value that can extract, be denoted as feature set F^N, wherein N is characterized The dimension of collection, N are related with the characteristic value number of species extracted and used；

S3.3: the feature set F obtained based on S3.2^N, using the method for semi-supervised learning to cross correlation results segment R_i[τ] is carried out Non line of sight identification.

6. a kind of acoustical signal non line of sight recognition methods based on semi-supervised learning according to claim 5, which is characterized in that

In S3.1, { Γ_a,Γ_τIndicate are as follows:

In S3.2, from { Γ_a,Γ_τIn extract characteristic value include: delay characteristics, wave character and Lai Si k-factor；

S3.3 includes the following steps:

S3.3.1: it respectively takes partial data as the monitoring data of known class respectively in sighting distance and non line of sight sample, is marked Label diffusion；

S3.3.2: Classification and Identification is carried out to the acoustical signal data in S3.3.1 after label is spread.

7. a kind of acoustical signal non line of sight recognition methods based on semi-supervised learning according to claim 6, which is characterized in that

S3.3.1's method particularly includes:

S3.3.1.1: setting label diffusion parameter is L；

S3.3.1.2: the acoustical signal data of each label known class and the acoustical signal data spacing of each unknown classification are calculated From distance calculation formula is as follows:

Wherein d is the distance between two acoustical signal data characteristics collection, and x, y are respectively the feature set of two acoustical signal data, and N is spy Dimension is collected, i is the index from 1 to N, x_iAnd y_iFor x, characteristic value of the y under currently index dimension；

S3.3.1.3: according to apart from calculated result, the acoustical signal data of all unknown classifications are ranked up from small to large；

S3.3.1.4: by the acoustical signal data apart from the smallest unknown classification, it is marked as classification mark identical with the given data Label.

8. a kind of acoustical signal non line of sight recognition methods based on semi-supervised learning according to claim 7, which is characterized in that

S3.3.2's method particularly includes:

S3.3.2.1: setting sorting parameter is K；

S3.3.2.2: distance between the acoustical signal data of each unknown classification and the acoustical signal data of each known class, institute are calculated It is identical as the calculation formula in S3.3.1.2 using calculation formula；

S3.3.2.3: to the acoustical signal data of each unknown classification, according to distance value calculated result, from small to large to all known The acoustical signal data of classification are ranked up；

S3.3.2.4: by the acoustical signal data of the smallest known class of distance value, the highest class label conduct of the frequency of occurrences The label of the unknown data.

Technical field

The invention belongs to the service technology fields based on indoor location, and in particular to a kind of sound letter based on semi-supervised learning Number non line of sight recognition methods.

Background technique

With popularizing for smart phone, the demand for services based on indoor location is increasing, such as indoor navigation, accurate battalion Pin, public safety etc., especially demand is bigger in the heavy constructions such as underground parking, market and museum.For needing above It asks, has proposed a variety of localization methods based on technologies such as sound, GSM, bluetooth, Wi-Fi, magnetic fields, and be based on localization of sound Technology has many advantages, such as and smart phone is completely compatible, positioning accuracy is high and at low cost, becomes most possible and solves in mobile phone room One of system of positioning.However, being managed from the point of view of the result of Microsoft's indoor positioning contest in 2018 and according to indoor ray acoustics By when path sighting distance (LOS) between sound source broadcasting equipment and receiving device is blocked, non line of sight (NLOS) phenomenon can be distance measurements It surveys and introduces a biggish non-minus deviation, as shown in Figure 1, the performance and stability of positioning system can be reduced.Non line of sight (NLOS) Phenomenon has become one of technical bottleneck of such technology, becomes the intelligent mobile terminal based on audio technology and applies in actual scene Huge challenge.

By identifying and abandoning NLOS measuring value, positioning accuracy can be improved merely with LOS measuring value, it can thus be concluded that non-view Accuracy away from identification becomes one of the determinant of indoor position accuracy.The now non line of sight recognition methods based on supervised learning What is used is the historical information of acoustical signal data, and when marked data volume is larger, the identification situation of non line of sight is preferable.But In practical applications, " label " information for obtaining a large amount of acoustical signal data is very difficult, needs to expend a large amount of manpower and material resources.This Problem limits application of the supervised learning method in the identification of acoustical signal non line of sight, and there is an urgent need to one kind can be based on a small amount of band The training data of label carries out non line of sight to a large amount of Unknown worm acoustical signal data and knows method for distinguishing.

Summary of the invention

For now with the technical problem in technology, the present invention provides a kind of non-views of the acoustical signal based on semi-supervised learning Away from recognition methods, the method for the present invention is according to the training data of a small amount of known class, a large amount of unknown acoustical signal data of automatic distinguishing, solution The actual application problem of acoustical signal of having determined non line of sight identification.

In order to solve the above technical problems, the present invention is resolved by the following technical programs:

A kind of acoustical signal non line of sight recognition methods based on semi-supervised learning, comprising the following steps:

S1: acquisition original sound signals x [n]；

S2: being detected and divided to the original sound signals x [n] acquired in S1, and cross correlation results segment R is obtained_i[τ]；

S3: to cross correlation results segment R obtained in S2_i[τ] carries out feature extraction and non line of sight identification.

Further, S2 includes the following steps:

S2.1: original sound signals x [n] is filtered and is enhanced, enhanced acoustical signal x'[n is obtained]；

S2.2: construction reference signal r [n], using reference signal r [n] to enhanced acoustical signal x'[n] carry out cross-correlation It calculates, obtains result R_x'r[τ]；

S2.3: to the result R obtained in S2.2_x'r[τ] is detected and is split extraction, obtains cross correlation results piece Section R_i[τ], the cross correlation results segment of i-th of enhanced acoustical signal of note are R_i[τ]。

Further, in S2.1, x'[n]=IFFT { FFT { x [n] } w [n] }, wherein w [n] is window function；

In S2.2,Wherein N be x'[n] length；

The specific method is as follows by S2.3:

To R_x'r[τ] carries out sequential detection, sets the length of sequential Load Signal segment as T_s, sequential Load Signal segment For seg [τ]=R_x'r[τ_s], wherein τ_s=[(i-1) T_s+1:iT_s]；Decision procedure in seg [τ] comprising useful signal is K { seg [τ] } >=thd, wherein thd is decision threshold, and K { } is that waveform kurtosis calculates symbol；If in seg [τ] including useful signal, Then the ID of sequential Load Signal segment and cross correlation results segment and beaconing nodes is carried out according to the broadcasting timeline of beaconing nodes Matching, is as a result denoted as a_i；The peak-peak position in cross correlation results segment is calculated, is denoted asIt cuts The subscript of acoustical signal and cross correlation results segment is taken to index are as follows:

Beaconing nodes a_iThe signal segment x ' of acoustical signal_i[n]=x'[idx_s:idx_e], cross correlation results segment R_i[τ]= R_x'r[idx_s:idx_e]。

Further, the window function is the compound window function of rectangular window and Blackman window composition, utilizes rectangular window LengthTo carry out bandpass filtering to original sound signals x [n].

Further, S3 includes the following steps:

S3.1: to cross correlation results segment R_iRelative gain-the time delay distribution of [τ] is estimated, { Γ is obtained_a,Γ_τ}；

S3.2: { the Γ obtained from S3.1_a,Γ_τIn extract the characteristic value that can extract, be denoted as feature set F^N, wherein N be The dimension of feature set, N are related with the characteristic value number of species extracted and used；

S3.3: the feature set F obtained based on S3.2^N, using the method for semi-supervised learning to cross correlation results segment R_i[τ] Carry out non line of sight identification.

Further, in S3.1, { Γ_a,Γ_τIndicate are as follows:

In S3.2, from { Γ_a,Γ_τIn extract characteristic value include: delay characteristics, wave character and Lai Si k-factor；

S3.3 includes the following steps:

S3.3.1: respectively taking partial data as the monitoring data of known class respectively in sighting distance and non line of sight sample, into Row label diffusion；

S3.3.2: Classification and Identification is carried out to the acoustical signal data in S3.3.1 after label is spread.

Further, S3.3.1 method particularly includes:

S3.3.1.1: setting label diffusion parameter is L；

S3.3.1.2: it calculates between the acoustical signal data of each label known class and the acoustical signal data of each unknown classification Distance, distance calculation formula are as follows:

Wherein d is the distance between two acoustical signal data characteristics collection, and x, y are respectively the feature set of two acoustical signal data, N It is characterized collection dimension, i is the index from 1 to N, x_iAnd y_iFor x, characteristic value of the y under currently index dimension；

S3.3.1.3: according to apart from calculated result, the acoustical signal data of all unknown classifications are ranked up from small to large；

S3.3.1.4: by the acoustical signal data apart from the smallest unknown classification, it is marked as class identical with the given data Distinguishing label.

Further, S3.3.2 method particularly includes:

S3.3.2.1: setting sorting parameter is K；

S3.3.2.2: the acoustical signal data of each unknown classification and the acoustical signal data spacing of each known class are calculated From used calculation formula is identical as the calculation formula in S3.3.1.2；

S3.3.2.3: to the acoustical signal data of each unknown classification, according to distance value calculated result, from small to large to all The acoustical signal data of known class are ranked up；

S3.3.2.4: by the acoustical signal data of the smallest known class of distance value, the highest class label of the frequency of occurrences Label as the unknown data.

Compared with prior art, the present invention at least has the advantages that the present invention is a kind of based on semi-supervised learning Acoustical signal non line of sight recognition methods, including data acquisition, the detection of acoustical signal and segmentation, feature extraction and non line of sight identification, energy Enough acquisitions have label acoustical signal data sample and without label data samples, and extract multiple features of acoustical signal data sample, It is then based on these characteristic values and carries out non line of sight identification using semi-supervised learning.It is existing to carry out non line of sight identification using supervised learning Method, this method is only when obtaining the acoustical signal data of a large amount of known class as training data, and recognition effect is more satisfactory, so And acquisition known class data are much more much bigger than obtaining unknown categorical data difficulty in practical application, the method for the present invention is obtaining After taking the acoustical signal data of a small amount of known class, the acoustical signal data of a large amount of unknown classifications are made full use of, the two is collectively as instruction Practice data and carry out non line of sight identification, saves manpower and material resources, and Classification and Identification effect is preferable, solve only a small amount of known sample In the case of the problem that identifies of acoustical signal non line of sight, the practical application for the indoor locating system based on audio technology provides the foundation.

To enable the above objects, features and advantages of the present invention to be clearer and more comprehensible, preferred embodiment is cited below particularly, and cooperate Appended attached drawing, is described in detail below.

Detailed description of the invention

In order to illustrate more clearly of the technical solution in the specific embodiment of the invention, specific embodiment will be retouched below Attached drawing needed in stating is briefly described, it should be apparent that, the accompanying drawings in the following description is some realities of the invention Mode is applied, it for those of ordinary skill in the art, without creative efforts, can also be attached according to these Figure obtains other attached drawings.

Fig. 1 is indoor acoustical signal sighting distance and non-line-of-sight propagation scene description；

Fig. 2 is data acquisition scenarios schematic diagram；

Fig. 3 is that the result after the method for the present invention identifies test data set collected is shown；

Fig. 4 by acquisition original sound signals image show.

Specific embodiment

In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with attached drawing to the present invention Technical solution be clearly and completely described, it is clear that described embodiments are some of the embodiments of the present invention, rather than Whole embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art are not making creative work premise Under every other embodiment obtained, shall fall within the protection scope of the present invention.

The embodiment of the present invention builds the indoor locating system based on audio technology using certain underground parking garage as experiment scene, into The data of row original sound signals acquire, and complete non line of sight identification, illustrate based on semi-supervised learning method in acoustical signal non line of sight Application in identification lays the foundation to further increase the indoor position accuracy based on audio technology.

A kind of acoustical signal non line of sight recognition methods based on semi-supervised learning of the present invention, comprising the following steps:

S1: acquisition original sound signals x [n], original sound signals are as sample data, including sighting distance and non line of sight sample This；It is illustrated in figure 4 some collected original sound signals image to show, the original signal includes 6 effective as shown in Figure 4 Signal segment is generated by 6 different beaconing nodes broadcast respectively；

The indoor locating system based on audio technology is built in signal scene as shown in Figure 2, whole system is by 6 beacon sections Point (number 1-6), 1 label composition, wherein beaconing nodes fixed height is the broadcast of 2.5 meters of progress acoustical signals, according to fixation Timing sends linear FM signal, i.e.,Wherein f₀For initial frequency, b₀ For chirp rate, b₀T is the time domain bandwidth of signal.And label carries out the reception of acoustical signal；

Selected data acquisition scene can be divided into 4 regions, wherein region 1 receives the sound letter of beaconing nodes 1,2,4,5 Number be line-of-sight signal, receive beaconing nodes 3,6 acoustical signal be non line of sight signal；Region 2 receives beaconing nodes 3,4,5 Acoustical signal is line-of-sight signal, and the acoustical signal for receiving beaconing nodes 1,2,6 is non line of sight signal；Region 3 receives beaconing nodes 1,2,3,4,5 acoustical signal is line-of-sight signal, and the acoustical signal for receiving beaconing nodes 6 is non line of sight signal；Region 4 receives letter The acoustical signal for marking node 6 is line-of-sight signal, and the acoustical signal for receiving beaconing nodes 1,2,3,4,5 is non line of sight signal；

Each region is approximately separated into and is made of the grid of 1m × 1m, grid intersection point is data collection point；

Customization label is mounted on tripod and adjusts height and is put in grid intersection point by region 1 to region 4 for 1.2m The acquisition of acoustical signal data is successively carried out, original sound signals are denoted as x [n].

S2: being detected and divided to the original sound signals x [n] acquired in S1, and cross correlation results segment R is obtained_i[τ], Specifically comprise the following steps:

S2.1: original sound signals x [n] is filtered and is enhanced, enhanced acoustical signal x'[n is obtained], pass through x'[n] =IFFT { FFT { x [n] } w [n] } is obtained, and wherein w [n] is window function, and window function is rectangular window and Blacknam in the present embodiment The compound window function of window composition, utilizes the length of rectangular windowTo carry out band logical to original sound signals x [n] Filtering；

S2.2: construction reference signal r [n], using reference signal r [n] to enhanced acoustical signal x'[n] carry out cross-correlation It calculates, obtains result R_x'r[τ],Wherein N be x'[n] length；

S2.3: to the result R obtained in S2.2_x'r[τ] is detected and is split extraction, obtains cross correlation results piece Section R_i[τ], the cross correlation results segment of i-th of enhanced acoustical signal of note are R_i[τ], method particularly includes:

To R_x'r[τ] carries out sequential detection, to determine the subscript call number of useful signal；Set sequential Load Signal segment Length be 50ms, be denoted as T_s=0.05f_s, sequential Load Signal segment is seg [τ]=R_x'r[τ_s], wherein τ_s=[(i-1) T_s+ 1:iT_s]；Decision procedure in so seg [τ] comprising useful signal is K { seg [τ] } >=thd, and wherein thd is decision threshold, K { } is that waveform kurtosis calculates symbol；If in seg [τ] including useful signal, the broadcasting timeline according to beaconing nodes is by sequential dress It carries signal segment and cross correlation results segment is matched with the ID of beaconing nodes, be as a result denoted as a_i；Calculate cross correlation results piece Peak-peak position in section, is denoted asIntercept the subscript rope of acoustical signal and cross correlation results segment It is cited as:

Beaconing nodes a_iThe signal segment x ' of acoustical signal_i[n]=x'[idx_s:idx_e], cross correlation results segment R_i[τ]= R_x'r[idx_s:idx_e], then successively intercept and store the acoustical signal segment and cross correlation results segment of all beaconing nodes；

S3: to cross correlation results segment R obtained in S2_i[τ] carry out feature extraction and non line of sight identification, specifically include as Lower step:

S3.1: to cross correlation results segment R_iRelative gain-the time delay distribution of [τ] is estimated, { Γ is obtained_a,Γ_τ, table It is shown as:

S3.2: { the Γ obtained from S3.1_a,Γ_τIn extract the characteristic value that can extract, be denoted as feature set F^N, wherein N be The dimension of feature set, N are related with the characteristic value number of species extracted and used；In the present embodiment, the characteristic value of extraction includes: Delay characteristics, wave character and Lai Si k-factor；

S3.3: the feature set F obtained based on S3.2^N, using the method for semi-supervised learning to cross correlation results segment R_i[τ] Non line of sight identification is carried out, is specifically comprised the following steps:

S3.3.1: respectively taking partial data respectively in sighting distance and non line of sight sample, (label diffusion parameter L is arranged in the embodiment =30, i.e., the sample of known sighting distance and non line of sight classification is respectively 30) monitoring data as known class, progress label expansion It dissipates, method particularly includes:

S3.3.1.1: setting label diffusion parameter is L=30；

S3.3.1.2: it calculates between the acoustical signal data of each label known class and the acoustical signal data of each unknown classification Distance, distance calculation formula are as follows:

Wherein d is the distance between two acoustical signal data characteristics collection, and x, y are respectively the feature set of two acoustical signal data, N It is characterized collection dimension, i is the index from 1 to N, x_iAnd y_iFor x, characteristic value of the y under currently index dimension；

S3.3.1.3: according to apart from calculated result, the acoustical signal data of all unknown classifications are ranked up from small to large；

S3.3.1.4: by the acoustical signal data apart from the smallest unknown classification, it is marked as class identical with the given data Distinguishing label；

S3.3.2: carrying out Classification and Identification to the acoustical signal data in S3.3.1 after label is spread, method particularly includes:

S3.3.2.1: setting sorting parameter is K=5；

S3.3.2.2: the acoustical signal data of each unknown classification and the acoustical signal data spacing of each known class are calculated From used calculation formula is identical as the calculation formula in S3.3.1.2；

S3.3.2.3: to the acoustical signal data of each unknown classification, according to distance value calculated result, from small to large to all The acoustical signal data of known class are ranked up；

S3.3.2.4: by the acoustical signal data of the smallest 5 known class of distance value, the highest classification mark of the frequency of occurrences Sign the label as the unknown data.

The result that the training set less for sample size, supervised learning and semi-supervised learning identify acoustical signal non line of sight As shown in Figure 3.Recognition result shows that proposed semi-supervised learning method can be in the acoustical signal for only obtaining a small amount of known class Under data cases, identification classification is carried out to a large amount of unknown classification acoustical signal data, classifying quality is better than supervised learning sorting algorithm. This method need not obtain a large amount of training datas, save manpower and material resources, solve acoustical signal in the case of only a small amount of known sample The problem of non line of sight identification.

Finally, it should be noted that embodiment described above, only a specific embodiment of the invention, to illustrate the present invention Technical solution, rather than its limitations, scope of protection of the present invention is not limited thereto, although with reference to the foregoing embodiments to this hair It is bright to be described in detail, those skilled in the art should understand that: anyone skilled in the art In the technical scope disclosed by the present invention, it can still modify to technical solution documented by previous embodiment or can be light It is readily conceivable that variation or equivalent replacement of some of the technical features；And these modifications, variation or replacement, do not make The essence of corresponding technical solution is detached from the spirit and scope of technical solution of the embodiment of the present invention, should all cover in protection of the invention Within the scope of.Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.