RU2523340C2 - Method for acoustic representation of spatial information for users - Google Patents

Abstract

FIELD: physics.

SUBSTANCE: method for acoustic representation of spatial information for users includes emitting a frequency-modulated ultrasonic pulse, binaural reception of an echo signal at two ultrasonic microphones, amplifying and converting electrical echo signals to acoustic signals, matched filtering of the echo signals, time-base expansion of the obtained responses in binaural reception channels, followed by perception of the signals by a human auditory analyser. The method includes preliminary testing for each channel of the auditory system of the user by presenting acoustic signals with different frequencies on a background of natural noise, performing statistical analysis of perception thereof by the auditory analyser of the user based on correct detection, missing and false detection of signals with respect to the total number of presented acoustic signals and obtaining adjusting gain coefficients for each channel. After matched filtering and time-base expansion the obtained responses in binaural reception channels, the method includes performing multichannel scaled-temporal filtering of responses via forward and inverse wavelet transform thereof, followed by weighted summation of output signals of each channel based on the obtained adjusting gain coefficients in each channel.

EFFECT: use of the invention improves recognition of objects by improving noise-immunity of the acoustic system.

6 dwg

Description

The present invention relates to the field of special acoustics and can be used for orientation of specialists whose professional activity is related to movement in conditions of limited visibility, for example, fighters of the Ministry of Emergencies in the fire, as well as for the rehabilitation of the visually impaired.

Known technical systems and devices designed for the visually impaired, allowing them to navigate in the environment. It is customary to single out two areas for the development of technical systems for visualizing the surrounding space to help the blind. The first direction includes instruments for indicating free path. They are the simplest and carry information only about the presence of an obstacle in the path of the blind (see China Patent CN 2907594, M.cl. A61F 9/08, published 06.06.2007, “Stereo ultrasonic device for helping the blind”, European patent EP 1025828, Mcl A61F 9/08, G01S 17/02, published 09.08.2000, German priority, “Assistance in orientation for the blind and visually impaired”; German patent DE 4212163, Mcl A61F 9/08 A61H 3 / 06, publ. 10/14/1993, "Optical echo system for the orientation of the blind"; Patent for the invention of the Russian Federation No. 2040234, M. CL A61F9 / 08, publ. 07/25/1995, "Ultrasonic locator for the blind"; Patent for the invention of the Russian Federation No. 2359287, M.cl. A61F 9/08, published on June 20, 2009. “Infrared Locator for Visually Impaired People”).

The most advanced device in this area is an Obstacle Detector for visually impaired people (see patent for the invention of the Russian Federation No. 2212871, Mcl A61F 9/08, published September 27, 2003), which contains a pulse generator, DAC, sound frequency generator and headset. The device emits radiation using a line of point infrared emitters and receives on a line of point photoconverters, which determine the direction of an obstacle at large viewing angles.

Devices that implement methods related to the first direction are the simplest, but carry information only about the presence of an obstacle in the path of the blind.

The second direction involves the use of hearing to the maximum extent possible for perception of the environment. Devices of this direction allow you to localize the object, get information about the direction and distance. In devices of this type, it is possible to organize scanning of several objects at the same time, and the sound signal carries information about the acoustic nature of objects. To ensure a wide field of view, binaural perception is used.

Close in technical essence to the present invention is a method of acoustic representation of spatial information (see Leslie Kay. Aerial sonars with acoustic representation of spatial information for the visually impaired. Animal Sonar Systems, edited by R. G. Busnel and J. F. Fish Plenum, New York, 1980).

The essence of the method of representing acoustic information is as follows. An ultrasound signal with linear frequency modulation (LFM) with a frequency range of about an octave is emitted into the medium. Binaural echo signals are received at two ultrasonic receivers. The received signals are multiplied with a reference signal to obtain a beat signal. Carry out low-pass filtering of the process. After amplification, the signals are sent to the head phones for perception by the human auditory system. The distance to the object is represented by the frequency of the audible signal, and the direction to the object is the interaural difference in the amplitudes of the signals falling on both ears. Thus, in the considered method, the multiplication of the echo signal and the reference chirp signal is performed (heterodyning operation). The oscillation of the frequency beat generated as a result of low-pass filtering carries information about the distance to the object.

Electronics and battery powered in a separate unit, placed in your pocket. This ensures that information is received about the direction and distance to the object (the farther it is, the higher the sound), as well as after training the user within certain limits about his shape, size and texture.

A device that implements this method is no longer a simple indicator of an obstacle, but to a certain extent an environmental analyzer with many important characteristics. However, the richer the information issued to the operator, the greater the psychoacoustic load he experiences. The operator processes a rather complex sound picture with the use of such parameters of the reflected signal as volume, pitch, timbre characteristics.

However, for the device to successfully perceive and classify the objects that make up the acoustic scene, a rather long learning process is required. At the same time, courses on mastering the apparatus in Germany last about a month.

In addition, in the above technical solution, the operation of heterodyning signals is selected as the procedure for matching the operating frequencies of the locator (40-120 kHz) with the frequency range of hearing, which leads to a change in the number of waves contained in the useful signal.

Similar methods include the method of orientation of the blind and the device for its implementation (see patent for the invention of the Russian Federation No. 2188611, IPC A61F 9/08, publ. 09/10/1999), which consists in the fact that optics are placed between the blind and the observed scene -electronic device with which to convert changes in scene illumination into proportional frequency-modulated sound signals through two channels and send them to the head phones of a blind person who judges the appearance, presence and disappearance of an object on the stage by changing the sound signal.

However, in this method, the optical signals do not carry information about the texture of the object.

The closest in technical essence to the proposed invention (prototype) is a method of acoustic presentation of spatial information for the visually impaired (see patent for the invention of the Russian Federation No. 2053746, IPC A61F 9/08, publ. 02/10/96). The essence of the method is as follows. An ultrasonic frequency modulated pulse is emitted. Binaural reception of the echo signal by two ultrasonic microphones is performed, amplifies and converts the electrical signals into acoustic ones, they carry out coordinated filtering of the echo signals, temporarily stretch the received responses in the binaural reception channels, followed by signal perception by the human auditory system.

This method ensures the coordination of the frequency range and duration of the acoustic signal with the auditory analyzer and obtaining complete information contained in the echo signal. At the same time, it should be noted that the main means of signal processing in the proposed method is a human auditory analyzer. Therefore, even a minimal violation of auditory functions leads to a deterioration in the ability of the operator to navigate in space, i.e. to reduce the recognition of objects.

Achievable technical result of the invention is to improve the recognition of objects that make up the acoustic scene, by selectively increasing the noise immunity of the system.

The achievement of the technical result is ensured by the fact that in the method of acoustic presentation of spatial information for users, which consists in emitting an ultrasonic frequency-modulated pulse, binaural receiving an echo signal by two ultrasonic microphones, amplifying and converting electric echo signals into acoustic, consistent filtering of echo signals , temporary stretching of received responses in binaural reception channels, subsequent perception of signals by a human auditory analyzer, Testing is carried out for each channel of the user's auditory system by presenting acoustic signals with different frequencies against a background of natural noise, performing a statistical analysis of their perception by the user's auditory analyzer, taking into account the correct detection, skipping and false detection of signals in relation to the total number of presented acoustic signals and obtaining corrective gain factors for each channel, and after the operation of matched filtering and time stretching The received responses in the binaural reception channels produce multichannel time-scale filtering of the responses by direct and reverse wavelet transform with subsequent weighted summation of the output signals of each channel, taking into account the received correction gains in each channel.

The influence of the distinctive features of the proposed method to achieve a technical result can be explained as follows.

The preliminary testing of the auditory system of the operator allows you to generate corrective time-scale gain factors for each channel, individual for each user. And when implementing multi-channel time-scale filtering of responses - direct and inverse wavelet transform (VP) - use the selected corrective gain in each frequency channel and perform the subsequent operation of summing the output signals of each channel.

Consider these operations in more detail. For the input process x (t) of an acoustic receiver, a two-dimensional representation in the time-scale domain is possible, obtained by applying a continuous wavelet transform [I. Dobeshi Ten lectures on wavelets. Izhevsk: Research Center "Regular and Chaotic Dynamics", 2001, 464 pp., Astafieva N.M. Wavelet Analysis: Fundamentals of the Theory and Application Examples. Advances in physical sciences. Volume 166, No. 11, 1996, pp. 1145-1170. Dremin I.M., Ivanov O.V., Nechitailo V.A. Wavelets and their use. Advances in physical sciences. Volume 171, No. 5, 2001, pp. 465-501. Dyakonov V.P. Wavelets. From theory to practice. M .: SOLON-R, 2002, 440 p. Kravchenko V.F., Rvachev V.A. “Wavelet” - systems and their application in signal processing. Foreign electronics. 1996, No. 4, pp. 3-20].

Continuous EP can be defined as the scalar product of the process x (t) under study and the basis functions ψ _ατ (t) [I. _Dobeshi Ten lectures on wavelets. Izhevsk: Research Center "Regular and chaotic dynamics", 2001, 464 p.]:

where the bar above indicates the operation of complex pairing.

The general principle of constructing the basis of the VP consists in the use of large-scale transformations with the compression parameter α and displacements with the shift parameter τ of the original wavelet function ψ (t) (the so-called mother wavelet):

должна удовлетворять еще одному условию:To be a wavelet, the basis functions ψ _ατ (t) ∈L ² (R) must have a number of necessary properties [6-10]. They should be: quadratically integrable, alternating (have a zero mean), while the wavelets should tend to zero by ± ∞, and for practical purposes - the faster, the better (and the wavelet should be well localized both in time and frequency) . In order to make the reverse EP possible, the wavelet spectral function

must satisfy one more condition:

The formula for continuous inverse wavelet transform is:

As can be seen from (4), the initial signal x (t) can be reconstructed through the integral sum of the same basis functions ψ _ατ (t) with weights in the form of the wavelet spectrum of the signal [W _ψ x] (α, τ). Here the constant C _ψ (3) acts as a normalizing coefficient similar to the coefficient (2π) ^1/2 normalizing the Fourier transform. The adjustable amplification factors selected in the preliminary testing in each frequency channel are weighting factors that determine the contribution of each of the scales to the total response received by the user's hearing aid. Then the operation is weighted summation in accordance with the ratio:

where W is the result of a time-scale transformation, γ _a are correction factors, a is the scale number, m = 0 ... N-1, N is the number of scale samples.

The proposed signal processing makes it possible to take into account the user's actual perception of acoustic signals and more fully transmit information about the environment to his auditory analyzer, which improves recognition of objects.

The invention is illustrated by drawings, in which Fig. 1 shows an example of a device for implementing the proposed method, Fig. 2 shows a block diagram of a direct wavelet transform algorithm using a Haar wavelet implemented in a Mathcad environment, Fig. 3 shows a block diagram of an algorithm the reverse wavelet transform performed in the MathCAD environment, Fig. 4 shows the probabilities of the correct detection of the presented implementations against the background of natural noise, Fig. 5 shows the structure of the control unit, and Fig. 6 shows a block diagram of al the operation algorithm of the statistical analysis unit.

A device that implements the processing in accordance with the considered method (Fig. 1) contains serially connected probe pulse generator 1, frequency modulation (FM) block 2, radiation path amplifier 3 and transmitter 4, right and left processing channels, each of which contains sequentially connected ultrasound transducer 5 (6), analog-to-digital transducer 7 (8) (ADC), matched filter (SF) 9 (10), memory block 11 (12), a set of time-scale filters 13 (14), adder 15 ( 16), digital-to-analog converter (DAC) 17 (18), an amplifier 19 (20), a headphone 21 (22), and the second input of the ADC 7 (8) of both channels is connected to the corresponding outputs of the clock generator 23, the corresponding inputs of the time-scale filters 13 (14) are connected to the corresponding outputs of the control unit 24, to the other corresponding outputs of which the inputs of the corresponding digital-to-analog converters 17 (18) are connected, and the memory blocks 11 (12), and the input of the control unit 24 is connected to the corresponding output of the clock generator 23.

The control unit 24 according to FIG. 5 comprises a series-connected unit 25 of statistical analysis, a random access memory 26 and a multiplexer 27, the outputs of which are used to connect to time-scale filters 13, 14, as well as a frequency divider 28, the outputs of which are used to connect to the corresponding the inputs of blocks 11 and 12 of the memory and the DAC 17, 18.

Consider the implementation of the proposed method in the above device.

. Например, при L=0,5 м Тсф $$\simeq \raisebox{1ex}{$2L$}\!\left/ \!\raisebox{-1ex}{$c$}\right.\sim 3мс.$$

После временного растяжения длительность сигнала, поступающего на слуховой аппарат пользователя (САП), составит Тсф~200 мс. Такое увеличение длительности важно при восприятии, поскольку сигналы длительностью меньше 0,2 с распознаются с трудом. Отклики СФ 9, 10, соответствующие импульсной характеристике рассеивающего объекта, записываются в блоки 11 (12) памяти. Реализации поступают на наборы масштабно-временных фильтров 13 (14), реализующих Вейвлет-преобразование, и производится их последующее суммирование в блоках 15 (16).After the radiation, the input implementation goes to the ADC 7, 8 and then to the digital SF 9, 10. Suppose that probing pulses with a carrier frequency of 80 kHz are used, and the average frequency of the audible range is 1.2 kHz. In this case, α ~ 66. If a signal of duration T is reflected from an object with a length in space L, the duration of the response of the SF Tf can be estimated as

. For example, at L = 0.5 m Tsf $$\simeq \raisebox{1ex}{$2L$}\!\left/ \!\raisebox{-1ex}{$c$}\right.\sim 3m\mathrm{from}.$$

After temporary stretching, the duration of the signal arriving at the user's hearing aid (SAP) will be Tsf ~ 200 ms. Such an increase in duration is important in perception, since signals with a duration of less than 0.2 s are hardly recognized. The SF responses 9, 10, corresponding to the impulse response of the scattering object, are recorded in the memory blocks 11 (12). Implementations are received on sets of time-scale filters 13 (14) that implement the wavelet transform, and their subsequent summation is performed in blocks 15 (16).

The operation of direct wavelet transform using the Haar wavelet is carried out in accordance with the expression:

- вейвлет-коэффициенты, соответствующие горизонтальному и вертикальному направлениям, для j-го масштаба; j=0, …, log₂N-1 - количество масштабов;

;

- количество строк и столбцов результирующей матрицы вейвлет-коэффициентов для j-го масштаба. Вейвлет Хаара реализован в среде Mathcad в качестве рабочего встроенного модуля (фиг.2).Where

- wavelet coefficients corresponding to horizontal and vertical directions for the j-th scale; j = 0, ..., log ₂ N-1 is the number of scales;

;

- the number of rows and columns of the resulting matrix of wavelet coefficients for the j-th scale. Wavelet Haar is implemented in the Mathcad environment as a working built-in module (figure 2).

In this case, the forecasting procedure for each of the scales of the time-scale plane is performed.

The operation of the inverse wavelet transform is performed in the MathCAD environment and is presented in the form of an algorithm (Fig. 3) in accordance with the expression:

where Ψ _{n, m} are basis functions.

The control unit 24 provides for reading information α times slower than writing and the arrival of correction factors in the time-scale filters 13 (14). Block 25 of the statistical analysis is designed to obtain corrective gain in each processing channel. Adjustable gains in each frequency channel are weighting factors that determine the contribution of each of the scales to the total response received by the user's auditory analyzer (SAP). In order to take into account the peculiarities of the SAP, it is necessary to initially take the values of the correction factors equal to unity. When conducting tests in the work of block 25 of the statistical analysis, the forced choice technique was used [Green D. Application of the theory of detection in psychophysics. TIIER 5, vol. 58, 1970], the essence of which is the choice of the subjects to decide on a certain time interval from several hypotheses. In our implementation example, there are 2 hypotheses: is there a signal or not a signal.

, где x - соответствует правильному обнаружению, y - общее количество предъявляемых стимулов. Полученные значения используются в качестве корректирующих коэффициентов. Для набора статистики оператору предъявляются сигналы с разными частотами на фоне естественных шумов, программным методом фиксируется статистика правильного обнаружения, пропуска сигналов и ложного обнаружения. Для набора репрезентативной выборки проводится необходимое количество испытаний. Пример кривых представлен на Фиг.4 (верхняя кривая - вероятность правильного обнаружения в устройстве для реализации предлагаемого способа, нижняя кривая - вероятность правильного обнаружения в прототипе). Введение коэффициентов позволяет повысить статистические показатели правильного обнаружения сигналов на фоне естественных шумов. Корректирующие коэффициенты берутся как величины, обратные к вероятности правильного обнаружения, для каждого масштаба, полученные в результате предварительного тестирования. Полученные значения корректирующих коэффициентов хранятся в ОЗУ 26 и с помощью мультиплексора 27 подаются в соответствующий канал масштабно-временных фильтров 13 (14). Делитель частоты 28 обеспечивает считывание информации в α раз медленнее записи из блоков 11 (12) памяти.As a result of the tests, detection statistics are collected. $$\raisebox{1ex}{$x$}\!\left/ \!\raisebox{-1ex}{$y$}\right.$$

, where x - corresponds to the correct detection, y - the total number of presented stimuli. The obtained values are used as correction factors. For statistics, the operator is presented with signals with different frequencies against a background of natural noise, the software records the statistics of correct detection, skipping signals and false detection. For a set of representative samples, the required number of tests is performed. An example of the curves is presented in Figure 4 (the upper curve is the probability of correct detection in the device for implementing the proposed method, the lower curve is the probability of correct detection in the prototype). The introduction of coefficients allows to increase the statistical indicators of the correct detection of signals against the background of natural noise. Correction factors are taken as the reciprocal of the probability of correct detection, for each scale, obtained as a result of preliminary testing. The obtained values of the correction factors are stored in RAM 26 and, using the multiplexer 27, are supplied to the corresponding channel of the time-scale filters 13 (14). Frequency divider 28 provides for reading information α times slower than recording from memory blocks 11 (12).

The received signals are converted into analog form using the DAC 17 (18) and, after amplification, they are transmitted to the headphones 21 (22).

The use of new operations compared to the prototype for the left and right processing channels:

- time-scale filtering of responses (wavelet transform) with correcting coefficients in each scale;

- the subsequent operation of summing the outputs of each channel of the time-scale filtering

allows you to get a new positive effect - more fully transmit information about the environment to the auditory analyzer operator (to increase the noise immunity of the method).

The authors conducted a preliminary assessment of the effectiveness of the proposed ^: method. A group of operators in the amount of three people was randomly presented with signals against a background of interference.

Before listening, a time-scale transformation operation was performed with the introduction of time-scale coefficients at each scale, followed by weight summation. Used signal frequencies: 100, 200, 400, 800, 1600, 3200, 6400 Hz. The upper graph (figure 4) corresponds to the probability of correct detection of the presented implementations using the proposed operations. The lower graph is based on the processing implemented in the prototype method. It can be seen that the gain lies in the range of 0.5-2 dB.

We show an example of the execution of the device blocks to implement the proposed method.

Sounding pulse generator 1, FM unit 2, radiation path amplifier 3, transmitter 4, ultrasonic transducers 5, 6, ADC 7, 8, matched filters 9, 10, memory blocks 11, 12, digital-to-analog converters 17, 18, clock generator 23 can be implemented similarly to the prototype. In particular, as matched filters 9, 10, digital or acoustoelectric compression filters can be used. In control unit 24 (FIG. 5), frequency divider 28 may be a binary counter. The frequency division coefficient is determined by the formula K = 2 ^a , where a is the counter capacity [L. Schegoleva, A. Davydov. Fundamentals of computer engineering and programming. L .: Energoizdat, 1981, from 144-145].

The multiplexer 27 (multiplex channel) serves for the exchange of information between RAM 26 and time-scale filters 13 (14) and is known in digital technology [Drozdov EA, Komarnitsky VA, Pyatibratov AP Multiprogramming digital computers. M .: Publishing. USSR Ministry of Defense, p. 378-383].

Block 25 of the statistical analysis is performed programmatically according to the block diagram of the algorithm (Fig.6).

Adders 15.16, RAM 26 are known in digital technology (L. Shchegoleva, A. Davydov. Fundamentals of computer engineering and programming. L.: Energoizdat, 1981 p., 155-201).

Type ADC 7, 8 is selected based on the required speed, determined by the frequency of the probe pulses. Since the maximum frequency used in such systems is 120 kHz, the necessary speed provides with a conversion time of less than 0.9 μs with a clock frequency in the range of 0.4-1.5 MHz. You can use the QMbox 15-48 series ADCs with a clock frequency of 1.5 MHz The temporal scaling of the signal is realized by analog-to-digital conversion, writing the samples of the signal into memory with one clock frequency, followed by reading from the memory and digital-to-analog conversion of another clock frequency, which is α times smaller, where α is the coefficient t transpose. The capacity of the memory blocks 11, 12 (PSU) should provide a record of reports of the input implementation. The number of samples is determined by N = f _d T _p , where T _p is the implementation duration, f _d is the sampling frequency. Storage devices are used in digital technology. DAC 17, 18 does not have special performance requirements, since the frequency of reading information from memory blocks is much lower than the recording frequency. Sets of time-scale filters that implement the wavelet transform are known in modern radio engineering (see patent for the invention of the Russian Federation No. 2246132, M.cl. G06F 17/14, published on July 20, 2004).

Claims (1)

The method of acoustic representation of spatial information for users, which consists in emitting an ultrasonic frequency-modulated pulse, binaural receiving an echo signal into two ultrasonic microphones, amplifying and converting electric echo signals into acoustic ones, coordinated filtering of echo signals, temporarily stretching the received responses in binaural channels reception, followed by the reception of signals by a human auditory analyzer, characterized in that it is pre-tested For each channel of the user's auditory system, by presenting acoustic signals with different frequencies against a background of natural noise, performing a statistical analysis of their perception by the user's auditory analyzer, taking into account the correct detection, skipping, and false detection of signals in relation to the total number of presented acoustic signals and obtaining corrective gain factors for each channel, and after the operation of coordinated filtering and temporary stretching of the received responses in the bin channels aural reception produce multi-channel time-scale filtering of responses by direct and reverse wavelet transform with subsequent weighted summation of the output signals of each channel, taking into account the received correction gains in each channel.

Images