WO2016107433A1 - Channel state selection method and device based on ternary coding - Google Patents

Abstract

Disclosed are a channel state selection method and device based on ternary coding. The method comprises: performing modulation processing on a sound source signal to generate (2L+1) level quantization signals x; performing mapping transformation on the quantization signals x to generate control signals p, m and z; performing selection processing on the control signals p, m and z and an M-channel feedback signal b, and then, outputting an M-channel state vector signal y; performing shaping processing on the state vector signal y, and then, generating the M-channel feedback signal b; and completing electric-acoustic conversion after the state vector signal y passes through a multi-channel power amplifier and an energy converter unit. The device comprises a modulator module, a mapping module, a selection module, a shaping module, and a multi-channel digital power amplifier and a loudspeaker array or a multi-voice-coil loudspeaker unit module, which are connected in sequence. The present invention can effectively reduce hardware resource overheads, reduce hardware power consumption, improve system stability, further improve the amplitude of an output signal, and improve conversion efficiency, and has a high signal-to-noise ratio output capacity.

Description

Channel state selection method and device based on three-state coding Technical field

The invention relates to a digital speaker coding method and device, in particular to a channel state selection method and device based on three-state coding.

Background technique

With the rapid development of VLSI manufacturing technology, the design and manufacture of the speaker system, the leading product of the electroacoustic industry, is gradually moving toward low power consumption, miniaturization and portability. In recent years, the semi-digital speaker system generated by the digital wave has successfully solved the power consumption and heat generation problems by using Pulse Width Modulation (PWM) Class D power amplifier driving technology, which greatly improved the whole process. The electroacoustic conversion efficiency of the system. However, the latter stage of the semi-digital speaker system still relies on a bulky LC low-pass analog filter to filter out the out-of-band high-frequency components of the digital pulse-modulated signal, and demodulate the modulated low-frequency envelope signal to complete Digital to analog conversion process. In order to eliminate the limitation of the analog LC filter, break through the digital bottleneck of the speaker unit, improve the integration level of the speaker system, and realize all digitization of all signal processing and transmission links of the speaker system, it is necessary to incorporate the speaker unit into the digital coding link to realize The digital coding of the speaker unit forms a digital speaker system, and finally the low-pass filtering characteristic of the speaker unit and the human ear structure completes the conversion of the digital encoding amount to the analog vibration amount, and moves the digital-to-analog conversion link to the electro-acoustic conversion physics. The stage is implemented, thereby eliminating the digital-to-analog conversion devices included in the conventional system and avoiding various electrical noises introduced by the digital-to-analog converter. Focusing on the core issue of digitization of speaker units, scholars at home and abroad have conducted extensive and in-depth theoretical and practical research on digital code modulation, digital power drive and digital speaker unit fabrication techniques in recent years. Thus, a new research field with digital speaker system design as the research direction has been formed.

In order to overcome the distortion introduced by PWM coding and ensure the high-fidelity playback of digital speakers, many experts, scholars and engineers have begun to develop a digital speaker system based on 1-bit delta-sigma coding, which is expected to be oversampled by delta-sigma modulation. The noise shaping technology pushes the system quantization noise power into the out-of-band high-frequency region to improve the sound quality level of the digital system. These digital speaker systems based on 1-bit delta-sigma encoding require only a simple low-pass filter to perform digital-to-analog conversion, and the hardware implementation is simple. At the same time, the system can transmit the desired audio in-band noise through oversampling and noise shaping techniques. In the high frequency area, high-fidelity sound quality is guaranteed. The digital system based on 1-bit delta-sigma modulation has the following shortcomings while having many of the above advantages: 1 is sensitive to clock jitter, and is easy to introduce nonlinear distortion due to clock jitter; 2 in order to maintain the stability of the modulation structure The allowable input signal has a small dynamic range; 3 requires a high switching rate, while the power MOSFET generates more nonlinear distortion components during the high-speed switching of the speaker load. It also causes an increase in heat generation, temperature rise, and efficiency reduction of the MOSFET.

In order to solve the defects of the 1-bit delta-sigma coding digital system, many scholars have turned to the research of digital systems based on multi-bit delta-sigma coding. Multi-bit delta-sigma modulation technology overcomes the above-mentioned shortcomings of 1-bit delta-sigma modulation, and it also has a fatal flaw itself - its modulation structure is between the frequency response of multiple speaker units (or voice coil units). The inconsistency and the degree of spatial position separation of a plurality of speaker units have high sensitivity, and it is easy to introduce a large coding error due to inconsistency of multiple unit frequency responses or separation of spatial positions.

In order to overcome the bias sensitivity defects of multi-bit delta-sigma modulation technology, since 1997, Professor Yasuda Akira of Japan’s Hosei University and Okamura’s engineer of Trigence Semiconductor have been working together to develop a digital system based on multi-bit delta-sigma coding. They proposed a system bias (frequency response and spatial position deviation) correction method based on dynamic mismatch shaping, and combined the Δ-∑ modulation and dynamic mismatch technology used in the system as “Dnote” technology; they will “Dnote” The technology implementation circuit is packaged into an IC chip, the "Dnote" chip, which uses the "Dnote" sample to create a variety of digital speaker system prototypes - 8-element piezoelectric linear array speaker system, 7-element piezoelectric ring array system and 6 The voice coil speaker system was exhibited at the 2008 Digital Audio Visual View. These systems do not require a power amplifier, LC filter, can be driven at a low voltage of 1.5V, and have directional control capabilities.

The dynamic mismatch shaper used in digital systems based on multi-bit delta-sigma modulation is essentially a multi-channel output state selection allocator based on various state selection strategies. The core idea of the dynamic mismatch shaping technique is to use a fast average to select each channel, and to push the signal error introduced by the system due to the deviation of each channel to the high frequency, thereby improving the signal to noise ratio in the audible sound frequency band. The three common dynamic mismatch shaping strategies are DWA (Data-Weighted Averaging), VFMS (Vector-Feedback mismatch-shaping), and TSMS (Tree-Structure mismatch shaping). Among them, the DWA selection strategy has the worst performance. The processed spectrum will still contain more obvious harmonic components at high frequencies, and DWA can only achieve first-order shaping. The shaping effect of TSMS and VFMS is better than that of DWA, and both TSMS and VFMS can achieve second-order. And the second-order or more shaping, the noise suppression capability of the VFMS is better than the TSMS in the same order. The traditional dynamic mismatch shaper (patent CN101803401A, patent CN102684700A, patent CN102239706A, patent CN102647191A) is designed for binary state coding. These traditional shapers can only contain two levels of "0" and "1". The binary coded signal of the state is subjected to shaping processing, and the three-state coded signal including the three level states of "-1", "0", and "1" cannot be directly shaped. Compared with the two-state coding, the biggest advantage of the three-state coding is that it can save the number of channels, reduce the amount of resources, and further improve the system integration. A three-state driving method is proposed in US Patent No. 2014169577, but the patent does not describe how to perform multi-channel three-state encoding selection, so that the influence of deviation between multiple channels on system signal-to-noise ratio degradation cannot be avoided.

The traditional three-state unit selection strategy is to divide the input signal into positive input and non-positive input, and then use the traditional two-state unit-based dynamic mismatch shaping selection strategy for the two inputs. There are minor changes on the basis of the design, and the structure is simple, but the signal-to-noise of the output signal after shaping is relatively low, and changes with the frequency. This tradition Although the three-state selection strategy reduces the amount of resources realized by the hardware, it also sacrifices the output signal-to-noise ratio performance. U.S. Patent No. 20120057727 proposes a strategy for selecting a tristate unit using channel averaging and time averaging, but this strategy has two drawbacks: (1) This method only gives a choice when the input signal approaches zero. The strategy does not give a selection strategy under other input conditions; (2) The time averaging strategy method will destroy the original signal and easily introduce harmonic distortion.

Aiming at the defects of the three-state selection strategy and combining the requirements of low power consumption, digitization and integration development, it is necessary to find a three-state selection method with excellent performance and simple implementation to achieve excellent channel deviation shaping effect.

Summary of the invention

The object of the present invention is to overcome the defects of the channel state selection strategy in the existing digital speaker system, and combines the requirements of low power consumption, digitization and integration development, and proposes a channel state selection method and device based on three-state coding. .

In order to achieve the above object, the technical solution adopted by the present invention is as follows:

A channel state selection method based on three-state coding includes the following steps:

1) The sound source signal is modulated to generate (2L+1) level-level quantized signals x;

2) the quantized signal x is transformed by mapping to generate control signals p, m and z;

3) control signal p, m, z and M channel feedback signal b is selected to generate an M channel state vector signal y;

4) The state vector signal y is shaped to generate a feedback signal b for the M channel update;

5) The state vector signal y is amplified by a three-state encoded multi-channel power amplifier to drive a plurality of speaker units in the speaker array or a plurality of voice coils in the multi-voice speaker, and the digital pattern is automatically completed by the speaker array or the multi-voice speaker Conversion and low pass filtering to convert the digitally encoded signal into an analog sound field signal.

Preferably, the method comprises the following steps:

1) The sound source signal is modulated to generate (2L+1) level-level quantized signals x;

2) the quantized signal x is transformed by mapping to generate control signals p, m and z;

3) The control signal p, m, z and the preset M channel feedback signal b is selected to generate an M channel state vector signal y; wherein the preset M channel feedback signal b can be selected as 0;

4) The state vector signal y is shaped to generate an updated M channel feedback signal b, and the control signals p, m, z and the updated M channel feedback signal b are selected to generate an updated M channel state vector signal. y;

5) The updated state vector signal y is amplified by a three-state encoded multi-channel power amplifier to drive a plurality of speaker units in the speaker array or a plurality of voice coils in the multi-voice speaker, and is automatically completed by a speaker array or a multi-voice speaker. Digital-to-analog conversion and low-pass filtering process convert the digitally encoded signal into an analog sound field signal.

Wherein, step 4) is repeated a plurality of times, and the M channel state vector signal y is updated multiple times until the M channel state vector signal y converges.

In the above technical solution, further, the (2L+1) level-level quantized signal x in the step 1), the quantized signal x Take any integer in the range [-L, L], where L is an integer and L ≥ 1.

In the above technical solution, further, the steps of the modulation processing in step 1) are as follows:

a) converting the source signal into a PCM coded signal having a sampling rate of f _s and a bit width of N;

b) processing the PCM coded signal with the sampling rate f _s and the bit width N by the upsampled low-pass filter to generate an oversampled PCM coded signal with a sampling rate f _o and a bit width of still N, where f _o = O _sr × f _s , O _sr is an oversampling factor;

c) the oversampled PCM coded signal is processed by multi-bit Δ-∑ modulation to generate a PCM coded signal x with a sampling rate still f _o and a quantization level level number (2L+1), wherein the quantization level level The quantity satisfies the condition: (2L+1)<2 ^N . The multi-bit delta-sigma modulation process is designed according to various multi-bit sigma-delta modulators - like the High-Order Single-Stage serial modulation method or multi-level (Multi-Stage (Cascade, MASH)) Parallel modulation method - the modulator structure and parameter design, the noise shaping process is performed on the oversampled signal outputted by the interpolation filter, and the noise energy is pushed to the area outside the audible band to ensure the modulated signal. There is a sufficiently high signal to noise ratio in the audible frequency band.

In the above technical solution, further, the control signals p, m, and z in the step 2) satisfy the condition: p+m+z=M, M≥L, where p represents the number of channels whose channel coding state is “1” , m represents the number of channels whose channel coding state is "-1", and z represents the number of channels whose channel coding state is "0".

In the above technical solution, further, the processing of the mapping transformation in step 2) is as follows:

When the quantized signal x ≥ 0, the values of the control signals p, m and z are as follows: p = x + floor ((Mx) / 2), m = Mpz, z = mod (Mx, 2); when the quantized signal x When <0, the values of the control signals p, m, and z are as follows: p=Mmz, m=-x+floor((M+x)/2), z=mod(M+x, 2). Where mod(Mx, 2) represents the remainder of (Mx) divided by 2, and floor((Mx)/2) represents the largest integer of the quotient not greater than (Mx) divided by 2, mod(M+x, 2) indicates ( M+x) divided by the remainder of 2, floor((M+x)/2): represents the largest integer of the quotient that is not greater than (M+x) divided by 2.

In the above technical solution, further, the expression of the M channel state vector signal y after the selection process in the step 3) is as follows:

y=[y ₁ ,y ₂ ,...,y _M ],

The state signal of the i-th channel is y _i , i ∈ {1, 2, ..., M}, y _{i is} selected from three state values of "-1", "0", "1" and satisfies

In the above technical solution, further, the steps of the selection process are as follows:

a) Set the feedback signal vector of the M channel to b=[b ₁ , b ₂ ,..., b _M ], where b _i represents the weight coefficient selected for the i th channel, and the feedback signal vector b is from large to small. Sort the order, generate the sorted feedback signal vector

among them

a weight coefficient representative of the selected i-th channel;

b) according to the weight coefficient of each channel after sorting

The position of the feedback signal vector after sorting

The output state of the channel corresponding to the first p larger value elements is set to "1", and the output state of the channel corresponding to the m smaller value elements is set to "-1", and the channel corresponding to the remaining z elements in the middle The output state is set to "0".

c) The M channel state vector signal y is generated by arranging all the above elements after the step b) selection processing in the order from 1 to M.

In the above technical solution, further, the shaping processing in the step 4) includes multi-channel filtering processing, and the transfer function of the filter used in each channel is

Where H(z) is the transfer function of the high-pass filter whose order is greater than one.

In the above technical solution, further, the step 2) the mapping transformation, the step 3) the selection processing, and the step 4) the shaping processing, the three steps jointly complete the shaping processing based on the three-state encoding, The output signal of the modulator module performs a mismatch shaping operation, which reduces the output signal-to-noise ratio degradation effect caused by the frequency response deviation between the subsequent channels. Based on the channel state selection method proposed by the present invention, the nonlinear distortion component of the synthesized signal introduced by the channel deviation is whitened, and the harmonic power at the specific frequency point is spread into the entire frequency band to be converted into a noise component. , eliminating the nonlinear distortion of the synthesized signal introduced by the harmonic component.

The tristate coding based shaping process is designed for the three-level state coded signal, which can save the algorithm hardware resources and save hardware power consumption. The shaping process of the three-state code is based on the performance of the number of times each channel has been recorded in the past, and determines which channels should be selected to participate in the sound signal restoration work at the current time. The shaping of the three-state code can optimize the combination of the channels participating in the signal restoration work to ensure that the total harmonic distortion of the synthesized acoustic signal is minimized. The shaping process of the three-state code is to control the state of the channel according to the criterion of the minimum harmonic distortion of the synthesized acoustic signal, and ensure that each channel participates in the synthesis of the acoustic signal according to the principle of equal probability. Each channel is optimal in its own response. Participate in the synthesis of acoustic signals in the state, thus ensuring the performance of the synthesized acoustic signals. Triangulation-based shaping is performed by averaging each channel as much as possible, equivalent to whitening the total harmonic components of the synthesized acoustic signal, dissipating these harmonic powers throughout the entire audio band, while using each unit as quickly as possible. After whitening, the noise is pushed to the high frequency, the in-band signal-to-noise ratio is improved, and the performance of the synthesized acoustic signal is improved.

In the above technical solution, further, the three-state coding in step 5) means that the output state of the channel at any time is only switched between three states of “1”, “0”, “-1”. . When the channel output state is "1", the input voltage on the speaker load terminal line is V _c . When the channel output state is "-1", the input voltage on the speaker load terminal line is -V _c , when the channel output state is " When 0”, the input voltage on the speaker load terminal line is 0, where V _c is the power supply voltage value of the power amplifier. The encoding and distribution of the multi-channel speaker load is completed by the multi-channel output state selection based on the three-state encoding, and the digital encoding and digitizing driving of each voice coil of the speaker array or the multi-voice speaker are realized.

The present invention also provides a channel state selection device based on three-state coding, which is characterized in that it comprises:

a modulator module modulating and encoding the sound source signal to generate a PCM coded signal x having a sampling rate f _o and a quantization level of 2L+1;

a mapping module, connected to the output of the modulator module, generating control signals p, m and z;

a selection module is connected to the output end of the mapping module, and is also connected to the input end and the output end of the shaping module to generate an M channel state vector signal y;

An shaping module is connected to the input end and the output end of the selection module to generate an M channel feedback signal b;

A multi-channel digital power amplifier and a speaker array or a multi-voice speaker unit module are connected to the output end of the selection module, and the power amplification of the M channel state vector signal y outputted by the selection module is completed by the multi-channel digital power amplifier for driving A plurality of voice coils in a speaker array or a plurality of voice coils in a multi-voice coil speaker are automatically digital-to-analog converted and low-pass filtered by a speaker array or a multi-voice speaker to convert the digitally encoded signal into an analog sound field signal.

Preferably, comprising:

a modulator module for modulating and encoding the sound source signal to generate a PCM coded signal x having a sampling rate f _o and a quantization level of 2L+1;

a mapping module, the input end of the mapping module is connected to the output end of the modulator module to generate control signals p, m and z;

a selection module, the input end of the selection module is connected to the output end of the mapping module, and generates an M channel state vector signal y according to the control signals p, m and z and the preset M channel feedback signal b;

An shaping module, an input end of the shaping module is connected to an output end of the selection module, an output end of the shaping module is connected to an input end of the connection module, and an updated version is generated according to the M channel state vector signal y M channel feedback signal b, the selection module generates an updated M channel state vector signal y according to the control signals p, m and z and the updated M channel feedback signal b;

a multi-channel digital power amplifier and a speaker array or a multi-speaker speaker unit module are connected to the output end of the selection module, and the power amplification of the updated M-channel state vector signal y outputted by the selection module is completed by the multi-channel digital power amplifier. Used to drive multiple speaker units in a speaker array or multiple voice coils in a multi-voice speaker, and automatically perform digital-to-analog conversion and low-pass filtering processing by a speaker array or a multi-voice speaker to convert digitally encoded signals into analog sound fields. signal.

In the above technical solution, the modulator module is composed of three modules: a format converter module, an interpolation filter module, and a multi-bit delta-sigma modulator module. The format converter module performs encoding format conversion on the sound source signal, and converts the sound source signal into a PCM encoded signal with a sampling rate of f _s and a bit width of N. The interpolation filter module performs up-sampling interpolation low-pass filtering on the output signal of the format converter module to generate an oversampled PCM coded signal with a sampling rate f _o and a bit width of N, where f _o =O _sr ×f _s , O _sr is an oversampling factor. The multi-bit delta-sigma modulator module performs multi-bit delta-sigma modulation on the output signal of the interpolation filter module to generate a PCM encoded signal x having a sampling rate of f _o and a quantization level of (2L+1).

Interpolation filter module, including at least 1 level FIR oversampling interpolation filter and 1 level CIC oversampling interpolation filter, FIR oversampling interpolation filter for smaller oversampling rate interpolation processing, CIC oversampling interpolation filter for large Oversampling rate Interpolation processing. The first stage uses the FIR oversampling interpolation filter, and the last stage uses the CIC oversampling interpolation filter.

In the above technical solution, the mapping module can be divided into offline and online implementations according to the offline calculation control signal and the online calculation control signal. The offline implementation has a fast processing speed but needs to occupy ROM resources. The online implementation consumes less hardware resources, but the processing speed is slightly slower than the offline mode.

A preferred implementation of the mapping module 2 - the specific process of the offline implementation is as follows:

In the offline implementation mode, the mapping module comprises a read only memory ROM1 module, a read only memory ROM2 module and a read only memory ROM3 module, and the values of the control signals p, m and z can be pre-calculated by an offline program and Pre-filled into the read-only memory ROM1 module, the read-only memory ROM2 module, and the read-only memory ROM3 module. The memory ROM1 module stores a control signal p, the address index is an input signal x of the mapping module; the memory ROM2 module stores a control signal m, the address index is an input signal x of the mapping module; the memory ROM3 module stores a control signal z, and the address index is The input signal x of the mapping module.

The contents of the read-only memory ROM1 module, the read-only memory ROM2 module, and the read-only memory ROM3 module are calculated by the following formula:

When x≥0, p=x+floor((M-x)/2), m=M-p-z, z=mod(M-x, 2);

When x<0, p=M-m-z, m=-x+floor((M+x)/2), z=mod(M+x, 2);

Among them, floor represents the smallest integer that is closest, and mod represents the remainder.

A preferred implementation of the mapping module - the specific process of the online implementation is as follows:

In the online implementation mode, the mapping module is composed of a symbol taking module, an absolute value module, a subtraction A1 module, a last bit module, a shift module, an addition module, a subtraction A2 module, a selector C1 module, and a selector C2 module. .

Taking the symbol bit of the input signal x of the symbol module output mapping module;

Taking the absolute value module output mapping module input signal x amplitude;

The subtraction A1 module subtracts the channel constant M from the output of the absolute value module, and outputs the difference;

Taking the last bit module to obtain the last bit code of the binary coded signal output by the subtraction A1 module as the output control signal z;

The shifting module shifts the output of the subtraction A1 module to the right by one bit, and outputs the shifted signal;

The adding module adds the output of the absolute value module and the output of the shifting module, and outputs an added signal;

The subtraction A2 module subtracts the channel constant M from the output of the last bit module, and subtracts the output of the addition module;

The first input signal of the selector C1 module is the output of the addition module, the second input signal is the output of the subtraction A2 module, the third input signal is the output of the symbolic module, and the output of the selector C1 module is the control signal p ;

The first input signal of the selector C2 module is the output of the subtraction A2 module, the second input signal is the output of the addition module, the third input signal is the output of the symbolic module, and the output of the selector C2 module is the control signal m. .

In the above technical solution, the selection module is composed of a sorting module, a positive selection module, a negative selection module, and an addition module. among them:

The sorting module sorts and outputs the input feedback signals b of the selection module in descending order.

The positive selection module sets the output state of the channel corresponding to the first p larger value elements of the sorting module output vector to "1", sets the output state of the channel corresponding to the remaining elements of the output vector to "0", and sets all channels. The output states are arranged in the order from 1 to M to generate an M channel state vector signal and output.

The negative selection module sets the output state of the channel corresponding to the m smaller value elements of the sorting module output vector to "-1", and sets the output state of the channel corresponding to the remaining elements of the output vector to "0", and all channels are The set output states are arranged in the order from 1 to M to generate an M channel state vector signal and output;

The addition module adds the channel state vector signals of the positive selection module and the negative selection module and outputs y.

In the above technical solution, the shaping module is composed of a subtraction module, a filter processing module, a minimum value search module, and an addition module.

The subtraction module subtracts the multi-channel state signal vector y of the shaping module from the multi-channel output feedback signal vector b of the shaping module, and outputs the multi-channel signal vector obtained by the subtraction process;

The filter processing module performs multi-channel filtering processing on the output signal vector of the subtraction module, and the transfer function of the filter used in each channel is H(z)-1, wherein H(z) is a high-pass filter with an order greater than one. Transfer function.

The minimum value search module receives the output of the filter processing module, and searches for the minimum value of the data transmitted on the multi-channel through multiple comparison processes, and outputs the opposite of the minimum value;

The adding module adds and outputs the output of the minimum value search module and the output of the filter processing module, and the output of the adding module is an updated feedback control signal b.

In the above technical solution, the multi-channel digital power amplifier and the speaker array or the multi-voice coil speaker unit module are composed of a multi-channel digital power amplifier and a transducer unit module having three-state driving capability.

The multi-channel digital power amplifier with three-state driving capability receives the M channel state vector signal y outputted by the selection module and performs multi-channel power amplification processing, and amplifies the status signals "-1", "0", and "1" sent by each channel. A power signal -V _c , 0 and V _c with drive capability is formed.

The transducer unit module receives an output signal of a multi-channel digital power amplifier having three-state driving capability and drives a plurality of voice coils in a speaker array or a plurality of voice coils in a multi-voice speaker, which is automatically replaced by a speaker array or a multi-voice speaker Digital-to-analog conversion and low-pass filtering are performed to convert the digitally encoded signal into an analog sound field signal.

The advantages of the present invention over the prior art are:

A. The invention adopts three-state coding, including three level formats of "-1", "0" and "1", under the same input condition, and only contains two kinds of electric energy such as "0" and "1" Compared with the traditional two-state coding in the flat state, the three-state coding method can reduce the number of channels by half, thereby effectively saving hardware resources occupied by the algorithm, reducing hardware resource overhead, saving hardware power consumption, and having good energy-saving characteristics, and is particularly suitable. Portable consumer electronics products can significantly improve the battery life of lithium battery powered products.

B. The channel state selection strategy based on three-state coding adopted by the invention, the method proposed by the invention can effectively improve the front-end input under the condition of the same number of channels, compared with the traditional channel state selection strategy based on two-state coding. The modulation depth of the signal enhances the stability of the system, further increases the amplitude of the output signal and improves the conversion efficiency.

C. The channel state selection strategy based on the three-state coding adopted by the invention white-processes the total harmonic components of the synthesized signal, disperses the harmonic power in the entire sound frequency band, and pushes the noise through the shaping means High frequency, improve the in-band signal-to-noise ratio, and also reduce the harmonic interference level of the unit splitter device, reduce the electromagnetic radiation level of the system, and reduce the interference of electromagnetic radiation to other surrounding electronic products. Compared with the traditional channel state selection strategy based on three-state coding, the proposed method can effectively optimize the multi-channel tri-state coding to ensure the total harmonic distortion of the synthesized signal is minimized, and the system can be significantly improved. Harmonic and noise attenuation suppression, improve system signal-to-noise ratio.

DRAWINGS

1 is a signal flow diagram of a channel state selection method based on three-state coding proposed by the present invention;

Figure 2 shows the feedback signal vector for sorting in the proposed method of the present invention

Schematic diagram of channel state selection;

3 is a block diagram showing an implementation of a channel state selection device based on three-state coding according to the present invention; wherein the selection module and the shaping module form a loop, and the output of the previous loop is updated, and can be looped multiple times until the M channel of the module output is selected. State vector signal y converges;

4 is a flow chart showing signal processing of a modulator module in a channel state selection method based on three-state coding proposed by the present invention;

FIG. 5 is a flowchart showing signal processing of an interpolation filter in a channel state selection method based on three-state coding according to the present invention; FIG.

6 is a schematic structural diagram of an offline implementation manner of a mapping module in a channel state selection method based on three-state coding according to the present invention;

7 is a schematic structural diagram of an online implementation manner of a mapping module in a method for selecting a channel state based on a three-state encoding according to the present invention;

8 is a structural diagram of a selection module in a channel state selection method based on three-state coding proposed by the present invention;

FIG. 9 is a structural diagram of a shaping module in a method for selecting a channel state based on a three-state encoding according to the present invention; FIG.

10 is a structural diagram of a multi-bit delta-sigma modulator used by a modulator module in a channel state selection method based on a three-state encoding according to the present invention;

FIG. 11 is a schematic diagram showing driving of a multi-channel digital power amplifier having three-state driving capability in a channel state selection method based on three-state encoding according to the present invention; FIG.

FIG. 12 is a graph showing the corresponding relationship between the single-channel output signal-to-noise ratio and the change of the input signal frequency and amplitude according to the channel state selection method based on the three-state coding proposed by the present invention and the channel state selection method based on the conventional VFMS improved algorithm. (a) a plot of the output signal-to-noise ratio as a function of the input signal frequency, and (b) a curve of the output signal-to-noise ratio as a function of the normalized input signal amplitude Line graph, wherein the input signal frequency is 5KHz;

Figure 13 shows the proposed single-channel output signal spectrum of the channel state selection method based on the three-state encoding and the traditional channel state selection method based on the improved VFMS algorithm, wherein the input frequency is 1 kHz;

detailed description

The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

The method and device for selecting channel state based on three-state coding proposed by the invention effectively reduces the hardware resource occupation of the algorithm, improves the stability of the system, and improves the harmonics by using the channel state selection method proposed by the invention. Noise suppression, while also increasing the maximum amplitude and conversion efficiency of the output signal.

A channel state selection device based on three-state coding according to the present invention is constructed, and the main body thereof is composed of a modulator module, a mapping module, a selection module, a shaping module, a multi-channel digital power amplifier and a speaker array or a multi-voice speaker unit module.

1) The modulator module comprises a format converter module, an interpolation filter module and a multi-bit delta-sigma modulator module, wherein the format converter module converts the sound source signal into a 16-bit, 48 KHz PCM coded signal output; the interpolation filter module presses The 2-level FIR interpolation filter and the 1-stage CIC interpolation filter have a total of 3 stages of filtering. The 16-bit, 48-KHz PCM coded signal output from the format converter is converted into a 16-bit, 3.072 MHz (48KHz x 64) PCM coded signal output. The first stage uses a 128-order FIR interpolation filter, the oversampling interpolation factor is 2, the second stage uses a 32-order FIR interpolation filter, the oversampling interpolation factor is 2, the third stage uses a CIC interpolation filter, and the oversampling factor is 16. The multi-bit delta-sigma modulator module converts the 16-bit, 3.072 MHz PCM encoded signal output by the interpolation filter into a 17-level, 3.072 MHz PCM encoded signal x output, as shown in Figure 10, using a delta-sigma modulator. It is a topology of 7th-order CIFB (Cascaded Integrators with Distributed Feedback), and its coefficients are shown in Table 1. Table 1 shows the parameter names and corresponding parameter values of the multi-bit delta-sigma modulator used by the modulator module in the channel state selection method based on the three-state coding proposed by the present invention.

Table 1

parameter name	A1	A2	A3	A4	A5	A6	A7
Numerical value	0.1955	0.1970	0.2032	0.2171	0.2429	0.2909	0.4204
parameter	B1	B2	B3	B4	B5	B6	B7	B8
Numerical value	0.1955	0.1970	0.2032	0.2171	0.2429	0.2909	0.4204	1
parameter	C1	C2	C3	C4	C5	C6	C7
Numerical value	0.0519	0.1121	0.1881	0.2945	0.4689	0.9021	3.0390

2) The mapping module converts the signal x output by the modulator module into control signals z, p and m outputs. The relationship between x and z, p, m is as follows, where the number of channels M is 8:

When x≥0, z=mod(M-x,2), p=x+floor((M-x)/2).,m=M-p-z;

When x<0, z=mod(M+x, 2), m=-x+floor((M+x)/2), p=M-m-z;

There are two main implementations of the mapping module. When the speed is the primary consideration, the implementation structure is shown in Figure 6. The relationship between the storage contents of the read-only memory ROM1, ROM2, and ROM3 and the input signal x is as shown in Table 2. Show. Resource occupancy When it is the primary consideration, its implementation structure is shown in Figure 7.

Table 2 shows the contents of the read-only memory ROM1 module, the read-only memory ROM2 module, and the read-only memory ROM3 module when the number of channels is 8.

Table 2

x	z	p		m
8	0	8	0
7	1	7	0
6	0	7	1
5	1	6	1
4	0	6	2
3	1	5	2
2	0	5	3
1	1	4	3
0	0	4	4
-1	1	3	4
-2	0	3	5
-3	1	2	5
-4	0	2	6
-5	1	1	6
-6	0	1	7
-7	1	0	7
-8	0	0	8

3) The selection module structure is shown in Figure 8, which mainly includes: a sorting module, a positive selection module, a negative selection module, and an addition module. Assuming that the signal x is 5, then z, p, and m are 1, 6, and 1, respectively, and assuming b is [10, 6, 7, 5, 4, 8, 9, 2], then the sorting module is pressed. The order from small to large is subscripted as [8,4,5,3,2,6,7,1], and the positive selection module subscripts and control signals p from small to large, the largest p in the first p of b The number corresponding channel is set to 1, the other channel is set to 0, the output is [1,1,1,1,0,1,1,0], and the negative selection module subscripts and control signal m according to the order from small to large, The channel corresponding to the m smallest number in b is set to -1, and the other channels are set to 0, the output is [0,0,0,0,0,0,0,-1], and the addition module is positively selecting the module and negative. Select module output addition, and its output signal y is [1,1,1,1,0,1,1,-1].

4) The shaping module structure is shown in Figure 9, which mainly includes: a subtraction module, a filter processing module, a minimum value search module, and an addition module. The transfer function of the filter is H(z)-1, where H(z) adopts a second-order filter structure with the expression (1-z ^-1 ) ² .

5) Multi-channel digital power amplifier and speaker array or multi-voice speaker unit module mainly consists of multi-channel digital power amplifier and transducer unit module with three-state driving capability, and multi-channel digital power amplifier with three-state driving capability is shown in Figure 11. Shown.

Example 1:

In the present embodiment, Table 3(a) shows the relationship between the output and the input of the channel state selection method based on the three-state coding proposed by the present invention in the 8-channel, and Table 3(b) also gives the traditional basis. The relationship between the output and the input of the channel state selection method of the improved algorithm of VFMS. The channel selection strategy of the traditional channel state selection method based on the improved VFMS algorithm is: (1) pre-order the M channel feedback signal b=[b ₁ , b ₂ ,..., b _M ]; (2) when x≥0 When the signal corresponding to the first x elements of the feedback signal b is set to "1", the channel corresponding to the other remaining elements is set to "0"; (3) when x < 0, the back -x of the feedback signal b is minimized. The channel corresponding to the element is set to "-1", and the channel corresponding to the other remaining elements is set to "0"; (4) The output signals of all channels are arranged in order from 1 to M to form the M channel output signal y. Assuming that the value of the feedback signal b is [10, 6, 7, 5, 4, 8, 9, 2], it can be seen from Table 3 that the present invention is compared to the conventional channel state selection method based on the improved VFMS algorithm. The proposed channel state selection method based on three-state coding uses more channels in the same time, the number of occurrences of the "0" coding state is greatly reduced, and "-1" or "1" can occur simultaneously in the same time. "Two encoding states, so you can use each output channel more quickly and evenly, pushing the noise to high frequencies and improving the in-band signal-to-noise ratio.

Table 3 shows a three-state code-based channel proposed by the present invention when the number of channels is 8, and the value of the feedback signal b is [10, 6, 7, 5, 4, 8, 9, 2] in the first embodiment of the present invention. The state selection method and the relationship between the output state vector y of the channel state selection method based on the VFMS improved algorithm and the input signal x.

table 3

(a)

(b)

Example 2:

In this embodiment, the correspondence between the single-channel SNR of the channel state selection method based on the three-state coding and the channel state selection method based on the improved VFMS algorithm is presented. The relationship, as shown in Figure 12, also shows the spectrum of the single-channel signal generated by the two methods when the input signal is 1 kHz and the input signal amplitude is -6 dB, as shown in Figure 13. . It can be seen from Fig. 12(a) that the single-channel signal-to-noise ratio of the traditional channel state selection method based on the improved VFMS algorithm increases with the increase of the input signal frequency, but the signal-to-noise ratio is low at all frequencies. According to the signal-to-noise ratio of the channel state selection method based on the three-state coding proposed by the present invention, the channel state selection method based on the three-state coding proposed by the present invention can also be seen from the spectrum diagram curve of FIG. The shaping and attenuation capabilities are superior to the traditional channel state selection method based on the improved VFMS algorithm. It can also be seen from Fig. 12(a) that the output signal-to-noise ratio of the proposed method does not substantially change with frequency and is a constant value. It can be seen from Fig. 12(b) that the output signal-to-noise ratio of the conventional channel state selection method based on the improved VFMS algorithm does not change linearly with the input signal amplitude, and the output signal-to-noise ratio of the proposed method is The amplitude of the input signal changes linearly. When the amplitude of the input signal is too large, the signal-to-noise ratio decreases because the shaping process is saturated, but the region is very narrow and the signal-to-noise ratio decreases less.

Finally, it should be noted that the above embodiments are merely illustrative of the technical solutions of the present invention and not limiting. While the invention has been described in detail herein with reference to the embodiments of the embodiments of the invention Within the scope of the claims.

Claims (18)

1. A channel state selection method based on three-state coding includes the following steps:

   1) The sound source signal is modulated to generate (2L+1) level-level quantized signals x;

   2) the quantized signal x is transformed by mapping to generate control signals p, m and z;

   3) control signal p, m, z and M channel feedback signal b is selected to generate an M channel state vector signal y;

   4) The state vector signal y is shaped to generate a feedback signal b of the M channel;

   5) The state vector signal y is amplified by a three-state encoded multi-channel power amplifier to drive a plurality of speaker units in the speaker array or a plurality of voice coils in the multi-voice speaker, and the digital pattern is automatically completed by the speaker array or the multi-voice speaker Conversion and low-pass filtering to convert the digitally encoded signal into an analog sound field signal.
2. The method for selecting a channel state based on three-state coding according to claim 1, comprising the steps of:

   1) The sound source signal is modulated to generate (2L+1) level-level quantized signals x;

   2) the quantized signal x is transformed by mapping to generate control signals p, m and z;

   3) the control signal p, m, z and the preset M channel feedback signal b is selected to generate an M channel state vector signal y;

   4) The state vector signal y is shaped to generate an updated M channel feedback signal b, and the control signals p, m, z and the updated M channel feedback signal b are selected to generate an updated M channel state vector signal. y;

   5) The updated state vector signal y is amplified by a three-state encoded multi-channel power amplifier to drive a plurality of speaker units in the speaker array or a plurality of voice coils in the multi-voice speaker, and is automatically completed by a speaker array or a multi-voice speaker. Digital-to-analog conversion and low-pass filtering process convert the digitally encoded signal into an analog sound field signal.
3. A channel state selection method based on three-state coding according to claim 1 or 2, characterized in that (2L+1) level-level quantized signals x in the step 1), the quantization The value of the signal x is any integer in the range [-L, L], where L is an integer and L ≥ 1.
4. The channel state selection method based on the three-state coding according to claim 1 or 2, wherein the step of the modulation processing in the step 1) is as follows:

   a) converting the source signal into a PCM coded signal having a sampling rate of f _s and a bit width of N;

   b) processing the PCM coded signal with a sampling rate f _s and a bit width N by an upsampled low-pass filter to generate an oversampled PCM coded signal with a sampling rate f _o and a bit width N, where f _o =O _sr ×f _s , O _sr is an oversampling factor;

   c) processing the oversampled PCM encoded signal by multi-bit delta-sigma modulation to generate a PCM encoded signal x having a sampling rate f _o and a quantization level level number (2L+1), said quantization level level The number of (2L+1)<2 ^N , the multi-bit delta-sigma modulation process is designed according to the design method of various multi-bit sigma-delta modulators, and the output of the interpolation filter is realized. The oversampled signal is subjected to noise shaping processing to push the noise energy into an area outside the audible band to ensure that the modulated signal has a sufficiently high signal to noise ratio in the audible frequency band.
5. The channel state selection method based on the three-state coding according to claim 1 or 2, wherein the control signals p, m and z in the step 2) satisfy the condition: p+m+z=M, M ≥ L, where p represents the number of channels whose channel coding state is "1", m represents the number of channels whose channel coding state is "-1", and z represents the number of channels whose channel coding state is "0".
6. The channel state selection method based on the three-state coding according to claim 1 or 2, wherein the processing of the mapping transformation in the step 2) is as follows:

   When the quantized signal x ≥ 0, the values of the control signals p, m and z are:

   p=x+floor((M-x)/2), m=M-p-z,z=mod(M-x,2);

   When the quantized signal x < 0, the control signals p, m and z are:

   p=M-m-z, m=-x+floor((M+x)/2), z=mod(M+x, 2);

   Among them, floor represents the smallest integer that is closest, and mod represents the remainder.
7. The channel state selection method based on the three-state coding according to claim 1 or 2, wherein the expression of the M channel state vector signal y after the selection process in the step 3) is as follows: y=[ y ₁ , y ₂ ,...,y _M ], wherein the state signal of the i-th channel is y _i , i ∈ {1, 2, ..., M}, y _i from "-1", "0", "1 "Three state values selected and satisfied

   
8. A channel state selection method based on three-state coding according to claim 1 or 2, wherein the step of selecting the processing is as follows:

   a) Set the feedback signal vector of the M channel to b = [b ₁ , b ₂ , ..., b _M ], where b _i represents the weight coefficient selected for the i th channel, i ∈ {1, 2, ..., M }, sorting the feedback signal vectors b in descending order to generate a sorted feedback signal vector

   

   among them

   

   a weight coefficient representative of the selected i-th channel;

   b) according to the weight coefficient of each channel after sorting

   

   The position of the feedback signal vector after sorting

   

   The output state of the channel corresponding to the first p larger value elements is set to "1", and the output state of the channel corresponding to the m smaller value elements is set to "-1", and the channel corresponding to the remaining z elements in the middle The output state is set to "0".

   c) generating all the above elements after the step b) selection processing in the order from channel 1 to M to generate the M channel State vector signal y.
9. The channel state selection method based on the three-state coding according to claim 1 or 2, wherein the shaping processing in the step 4) comprises multi-channel filtering processing, and the transfer function of the filter used in each channel is for

   

   Where H(z) is the transfer function of the high-pass filter whose order is greater than one.
10. The channel state selection method based on the three-state coding according to claim 1 or 2, wherein the three-state coding in the step 5) means that the output state of the channel at any time is only "1". Switch between the three states “0” and “-1”; when the channel output state is “1”, the input voltage on the speaker load terminal line is V _c ; when the channel output state is “-1”, The input voltage on the speaker load terminal line is -V _c ; when the channel output state is "0", the input voltage on the speaker load terminal line is 0; where V _c is the power supply voltage value of the multi-channel power amplifier, based on three The multi-channel output state of the state code selects the code distribution of the multi-channel speaker load, and digitally encodes and digitizes the voice coils of each array element or multi-voice speaker of the speaker array.
11. A channel state selection device based on three-state coding, comprising:

    a modulator module for modulating and encoding the sound source signal to generate a PCM coded signal x having a sampling rate f _o and a quantization level of 2L+1;

    a mapping module, connected to the output of the modulator module, generating control signals p, m and z;

    a selection module is connected to the output end of the mapping module, and is also connected to the input end and the output end of the shaping module 4 to generate an M channel state vector signal y;

    An shaping module is connected to the input end and the output end of the selection module to generate an M channel feedback signal b;

    A multi-channel digital power amplifier and a speaker array or a multi-voice speaker unit module are connected to the output end of the selection module, and the power amplification of the M channel state vector signal y outputted by the selection module is completed by the multi-channel digital power amplifier for driving A plurality of voice coils in a speaker array or a plurality of voice coils in a multi-voice coil speaker are automatically digital-to-analog converted and low-pass filtered by a speaker array or a multi-voice speaker to convert the digitally encoded signal into an analog sound field signal.
12. The channel state selection device based on the three-state coding according to claim 11, comprising:

    a modulator module for modulating and encoding the sound source signal to generate a PCM coded signal x having a sampling rate f _o and a quantization level of 2L+1;

    a mapping module, the input end of the mapping module is connected to the output end of the modulator module to generate control signals p, m and z;

    a selection module, the input end of the selection module is connected to the output end of the mapping module, and generates an M channel state vector signal y according to the control signals p, m and z and the preset M channel feedback signal b;

    An shaping module, an input end of the shaping module is connected to an output end of the selection module, an output end of the shaping module is connected to an input end of the connection module, and an updated version is generated according to the M channel state vector signal y M channel feedback signal b, the selection module generates an updated M channel state vector signal y according to the control signals p, m and z and the updated M channel feedback signal b;

    a multi-channel digital power amplifier and a speaker array or a multi-speaker speaker unit module are connected to the output end of the selection module, and the power amplification of the updated M-channel state vector signal y outputted by the selection module is completed by the multi-channel digital power amplifier. Used to drive multiple speaker units in a speaker array or multiple voice coils in a multi-voice speaker, and automatically perform digital-to-analog conversion and low-pass filtering processing by a speaker array or a multi-voice speaker to convert digitally encoded signals into analog sound fields. signal.
13. A channel state selection device based on three-state coding according to claim 11 or 12, wherein the modulator module comprises a format converter module, an interpolation filter module and a multi-bit delta-sigma modulator module ,

    The format converter module is configured to perform encoding format conversion on the sound source signal, and convert the sound source signal into a PCM coded signal with a sampling rate of f _s and a bit width of N.

    The interpolation filter module is configured to perform up-sampling interpolation low-pass filtering on the output signal of the format converter module to generate an oversampled PCM coded signal with a sampling rate f _o and a bit width N, where f _o =O _sr × f _s , O _sr is an oversampling factor,

    The multi-bit delta-sigma modulator module is configured to perform multi-bit delta-sigma modulation processing on the output signal of the interpolation filter module to generate a PCM encoded signal x having a sampling rate f _o and a quantization level of (2L+1).
14. The channel state selection device based on the three-state coding according to claim 11 or 12, wherein the offline implementation of the mapping module is as follows:

    In the offline implementation mode, the mapping module includes a read only memory ROM1 module, a read only memory ROM2 module, and a read only memory ROM3 module, and the values of the control signals p, m, and z are pre-calculated by an offline program and pre-filled separately. In the read only memory ROM1 module, the read only memory ROM2 module and the read only memory ROM3 module, the memory ROM1 module stores the control signal p, the address index is the input signal x of the mapping module; the memory ROM2 module stores the control signal m The address index is the input signal x of the mapping module 2; the memory ROM3 module stores the control signal z, and the address index is the input signal x of the mapping module 2,

    The contents of the read-only memory ROM1 module, the read-only memory ROM2 module, and the read-only memory ROM3 module are calculated by the following formula:

    When x≥0, p=x+floor((M-x)/2), m=M-p-z, z=mod(M-x, 2);

    When x<0, p=M-m-z, m=-x+floor((M+x)/2), z=mod(M+x, 2);

    Among them, floor represents the smallest integer that is closest, and mod represents the remainder.
15. The channel state selection device based on the three-state coding according to claim 11 or 12, wherein the online implementation manner of the mapping module is as follows:

    In the online implementation mode, the mapping module includes a symbol module, an absolute value module, a subtraction A1 module, and a fetching module. Last module, shift module, addition module, subtraction A2 module, selector C1 module and selector C2 module,

    Taking a symbol module for outputting a sign bit of the mapping module input signal x;

    Taking an absolute value module for outputting the amplitude of the mapping module input signal x;

    The subtraction A1 module is configured to subtract the channel number M from the output of the absolute value module and output the binary coded signal of the difference;

    Taking the last bit module, used to obtain the last bit code of the binary coded signal output by the subtraction A1 module, as the output control signal z;

    The shifting module is configured to output the signal after shifting the output of the subtraction A1 module to the right by one bit;

    The adding module adds the output of the absolute value module and the output of the shifting module, and then adds the signal output;

    The subtraction A2 module is used to subtract the channel number M from the output of the last bit module, and then subtract the output of the addition module;

    The first input signal of the selector C1 module is the output of the addition module, the second input signal is the output of the subtraction A2 module, the third input signal is the output of the symbolic module, and the output of the selector C1 module is the control signal p ;

    The first input signal of the selector C2 module is the output of the subtraction A2 module, the second input signal is the output of the addition module, the third input signal is the output of the symbolic module, and the output of the selector C2 module is the control signal m. .
16. The channel state selection device based on the three-state coding according to claim 11 or 12, wherein the selection module comprises a sorting module, a positive selection module, a negative selection module, and an addition module, wherein:

    a sorting module, configured to sort and output the input feedback signals b of the selection module in descending order;

    Positive selection module, which is used to set the output state of the channel corresponding to the first p larger value elements of the sorting module output vector to "1", and set the output state of the channel corresponding to the remaining elements of the output vector to "0", all The output state set by the channel is arranged in the order from 1 to M to generate an M channel state vector signal and output;

    The negative selection module is configured to set the output state of the channel corresponding to the m smaller value elements of the sorting module output vector to “-1”, and set the output state of the channel corresponding to the remaining elements of the output vector to “0”, The output states set by all channels are arranged in the order from 1 to M to generate an M channel state vector signal and output;

    An adding module is configured to add channel state vector signals of the positive selection module and the negative selection module and output a state vector signal y.
17. The channel state selection device based on the three-state coding according to claim 11 or 12, wherein the shaping module comprises a subtraction module, a filter processing module, a minimum value search module, and an addition module.

    a subtraction module, configured to subtract the multi-channel feedback signal vector b of the shaping module from the input multi-channel state signal vector y, and output the multi-channel signal vector obtained by the subtraction process;

    The filter processing module performs multi-channel filtering processing on the output signal vector of the subtraction module, and the transfer function of the filter used in each channel is H(z)-1, wherein H(z) is a high-pass filter with an order greater than one. Transfer function

    The minimum value search module receives the output of the filter processing module, and searches for the multi-channel transmission through multiple comparison processing. Send the minimum value of the data and output the opposite of the minimum value;

    The adding module adds and outputs the output of the minimum value search module and the output of the filter processing module, and the output of the adding module is an updated feedback control signal b.
18. A channel state selection device based on three-state coding according to claim 11 or 12, wherein said multi-channel digital power amplifier and speaker array or multi-voice speaker unit module comprises multi-channel with three-state driving capability Digital power amplifier and transducer unit module;

    The multi-channel digital power amplifier with three-state driving capability is used for receiving the updated M-channel state vector signal y outputted by the selection module and performing multi-channel power amplification processing, and the status signals "-1" and "0" sent by each channel are And "1" are amplified to form a power signal with drive capability -V _c , 0 and V _c ;

    a transducer unit module for receiving an output signal of a multi-channel digital power amplifier having three-state driving capability and driving a plurality of voice coils in a speaker array or a plurality of voice coils in a multi-voice speaker, by a speaker array or a multi-tone The ring speaker automatically performs digital-to-analog conversion and low-pass filtering to convert the digitally encoded signal into an analog sound field signal.

Images