TW556153B - Fast adaptive differential pulse coding modulation method for random access and channel noise resistance - Google Patents

Abstract

The present invention is a fast adaptive differential pulse coding modulation method for random access and channel noise resistance, which uses a modulation mechanism for encoding. During encoding, it starts the decoding from an arbitrary location in the compressed audio file, and it will not diverge when the compressed audio file is subjected a certain level of interference during the transmission. After decoding, the audio source can be recovered to a clear level. The method uses only the adder and shifter to complete the mechanism without using the multiplier. Thus, the present invention has the advantages of simple circuit and lower cost comparing to the conventional method.

Description

A7 A7 V. Description of the invention (1) Technical field of the invention ... Ready-to-read and anti-channel noise fast adaptation differential ship code Japanese Shishi Shibu 'especially a kind of use-tuning structure for coding, solution : σ sacred contraction begins to decode at any position, and after compressing the end of the sound, the sound source can still be restored to a clear level. ^ 于 解 ^ (二) Background of the invention: Press the 'technical wheel to never stop, and peak _Speed continues to progress, especially the development of the electronics industry is more so-called._ Qian Qianli, the development of the past five years is more than the achievements of the past decades, and almost all electronic products have reached the goal of smart, digital system in addition to In addition to being more stable by analogy, and with a more expected design, regardless of the performance of the design or the results presented, the digital mode is better than the analog system. The traditional technology of processing voice signals Shooting, ambiguity, and digital processing technology, analog processing technology is an older custom technology, so I won't go into details here, but only explain the digital processing mode. For the voice technology of ADPCM, please refer to two patent cases, one is the invention patent of "ADPCM sound generator" of the Republic of China Patent Gazette Announcement No. 9243, and the other is the Republic of China Patent Gazette Announcement No. 41. No. 32, No. 3, "Voice coding fast adaptation differential pulse code modulation method, the invention patent, the predictor used by the decoder of the second patent case (Predictor) only uses unit delay (Unit Delay) to achieve, If there is any channel noise or random capture, the decoded audio source signal may drift. This paper size applies the Chinese National Standard (CNS) A4 specification (210 X 297 ^^ 1)

φ Μ -------- ^ --------- (Please read the notes on the back before filling out this page) 556153 Printed by the Consumers ’Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs A7 V. Invention Description ( >) The elephant 'is even worse for the divergence of the audio signal. In addition, CCITT G.726's adaptive differential pulse code modulation method (lobe grid's although it has better anti-channel noise and random access mechanisms, but the basin mine path is too complicated (including wire addition, limb multiplication_, The relative increase is increased, and the present invention only uses the shifter to complete the mechanism. Compared with the present invention, it has the advantages of a simple circuit and a lower cost. Using the technology of the present invention, __ ( (Reqing nng Time) can be shortened, which can greatly reduce the time for decompression. (3) Brief description of the invention ... * In order to solve the above problems, the inventor has studied and researched for a long time, and has continuously developed and tested. Therefore, the present invention was obtained to solve the conventional technology. In the conventional technique, the ^^ axis, _E axis (examination) or decoding of the narration job title with _L may produce drifting wires, and even more, the sound of the sign = divergence phenomenon disappears, or is aimed at The historical shrinking circuit of the foregoing problem is expected to increase the cost because it is too complicated. The present invention is a method and mechanism for effectively solving the problem. The main purpose f of the tile H code is to use a modulation mechanism to perform the shrinking. Richtek can be used for compressed audio broadcasting The position starts to decode, and when the compression block is transmitted to a certain degree during the transmission, the sound source can still be restored to-clearness. It is not necessary for the government to use ΓΓ sun and moon's other "quotation" is to provide- Simple modulation mechanism, the mechanism can be completed by using only theft and shifter, so the present invention has a simple circuit and a lower cost of H. Using the paper method t paper standard applicable _i ^ TNS) A4i (21〇 x 2ir—- (Please read the precautions on the back before filling out this page} Μ -------- ^ --------- ^ 9 !. 556153

V. Description of invention (j) In order to further understand the key points and contents of the technology of this case, the review committee will use the following drawings, illustrations and detailed descriptions of inventions to help the review work. Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs (4) Schematic description of the invention: Figure 1 is a compressed flowchart of the invention. FIG. 2 is a decompression flowchart of the present invention. Figure 3 is a schematic diagram of the hardware architecture of the system of the present invention. Figure 4 is a schematic diagram of the internal architecture of the ADPCM hardware of the present invention. Explanation of drawing number: 11 input the sound source 12 parameter initialization 13 calculate the difference dif of the sound source X (k) and the predicted value pred (initial value is zero) 14 if dif is greater than or equal to zero, sign = 0, otherwise sign = 8, then the dif Take the absolute value (vdif) 15 and calculate it by the continuous approximation method △ 151 vdif > = step? 152 Δ = Δ | temp vdif = vdif-step vpdi ff = vpdi f f + step 153 step = step »l temp = temp» l _: __ 4 This paper size applies to China National Standard (CNS) A4 (210 x 297 male dreams) 1 ^ .Packing -------- Order --------- AVI (Please read the back first Please pay attention to this page and fill in this page again) 556153 Printed by A7 _B7___, Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 5. Description of the Invention (4) 154 temp = 0? 16 Use the adder to get the quantized signal pred 161 sr_kl = pred + / —vpdiff 162 i = 0 163 al & (0x01 «i) = 1? 164 pred = pred + sr__kl 165 srjd = sr—kl« li = i + l

166 i = 7? S 17 Update the value of al according to the values of b and sign, and limit al to a value between 0 and 127. b is an adjustment constant. 18 Look up the table according to Δ and index. If sign = 0, pred = pred + p, otherwise pred = pred-p; according to the lookup table, update the values of index and step 19 compressed data δ = Δ I sign 21 input compressed data δ 22 take the sign of δ and the size value Δ 23 indexTable table according to The change value is used to update the value of index, and the initial value of a is set to 4-24. vpdiff 241 (delta & a) = 1? 242 vpd iff = vpd if f + step 243 step = step »l 244 a 》 1 _5_ This paper size applies to China National Standard (CNS) A4 (210 x 297 mm) ----------- Aw- ^ -------- ^ ----- ---- (Please read the precautions on the back before filling out this page) 556153 A7 B7 V. Description of the invention (ί) 245 a = 0? 25 Use the loop method to get the quantized signal pred 251 sr_kl = pred + /-vpdiff 252 i = 0 253 al & (0x01 «i) = i? 254 pred = pred + sr__kl 255 sr—kl = sr—kl < < li = i + l 256 i = 7 ?, 26 according to b and The value of sign updates the ai value and limits ai to a value between 0 and 127. , B is an adjustment constant 27. According to the look-up table, the value of step 28 is updated. 28 Val is the decompressed sound source, which is output 31 CPU (central processing unit) 32 ADPCM Engine (ADPCM engine) 321 ADD / SUB & SHIFT (calculation Processor) 322 REGISTER 323 ADPCM State 324 ADDRESS GENERATOR 33 ROM (Read Only Memory) 34 RAM (Random Access Memory) 35 Analog / Digital Converter ( Analog to Digital Converter (ADC) 36 Analog to Digital Converter (DAC) (5) Detailed description of the invention. The preferred embodiment of this creation will be described with drawings. This paper size is in accordance with Chinese National Standard (CNS) A4 (210 X 297 mm) (please read the precautions on the back before filling this page) 吁 w MM a ··· am a so, 1 ϋ n tm— ϋ ϋ I · Printed by the Consumers' Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs 556153 A7 B7 V. Description of the invention (G) Refer to Figure 1, which is the compression process of the sound source of the present invention in PCM format. The steps are as follows: 11: The sound source turns in. The sound source is a sampled sound source in PCM format, and its value is represented by X (k). 12 ·· Parameter initialization, restore the index (index value), pred (default value), △ (size value) to the initial value (that is, zero). The value of temp is set to 4 for △ And the setting of 3 times. 13 · δ The difference dif between the meaning of the ten different sound sources (k) and the predicted value pre (j (the initial value is zero)), the formula is as follows: dif = X (k)-pred 14. If dif is greater than or equal to zero, otherwise sign = 8, then take the absolute value of dif), the formula is as follows: sign = sgn (dif) vdif = abs (dif) where sgn is a sign function with a data length of 4 bits. If it is negative, the output value is 8, otherwise it is equal to 0. abs takes the absolute value of the function in parentheses. 15: △ is calculated by the continuous approximation method, and the formula of the continuous approximation method includes the following: 151: vdif > = step? (Is the dif of the absolute value greater than or equal to the step value), and if the result is (Yes), then execute Step 152; if the result is No, skip step M? This paper size applies the Chinese National Standard (CNS) A4 specification (21 × 297 mm 556 153. Printed by the Consumers ’Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs A7 _____B7___ _ 5. Explanation of the invention (;?) Execute the step 153 directly. 152: A = A | temp (the judgment value of Δ and temp or OR) vdif = vdif-step vpdiff = vpdiff + step 153 ·· step = step »l temp = temp» l 154: temp = 0 ?, if the result is Yes, proceed to step 16; if the result is No, then jump back to 151 and perform continuous approximation again Method 16: Use the loop method to obtain the quantized signal pre (j) using the adder. The formula used includes the following: 161: sr—kl = pred +/— vpdiff 162: i = 0 163: al & ( 0x01 «i) = 1 ?; if the execution result is Yes, execute 164, and if the result is No, execute 165 164 pred = pred + sr_kl 165: sr—kl = sr—kl «li = i + l, 166 ·· i = 7 ?, if the execution result is yes (when Yesy, then execute the next step 17 'the result is negative ( No.), execute M3. 17: Update the value of al according to the values of b and sign, and limit al to a value between 0 and 127. b is an adjustment constant. Its formula is: pred = (pred + (al »l )) »7 This paper size is applicable to China National Standard (CNS) A4 specification (210 x 297 mm ^ (Please read the precautions on the back before filling out this page) I Loading -------- Order --- ------ 556153 Printed A7 by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs ——----_____ B7 —__ 5. Description of the invention (?) ^ '1 al = al +/- b in steps 16-17 It is a modulation mechanism. The present invention is to use the modulation mechanism to perform encoding. When decoding, decoding can be started at any position in the compressed sound file, and when the compressed sound groove is subject to a certain degree of interference during transmission, As for divergence, after decoding, the sound source can still be restored to a clear level. 18: According to the table lookup, update the values of index and step, the formula is: index = index + indexTable ( Δ) step = stepTable (index) 19: Compressed data δ = Δ I sign, where △ is represented as a size value, and an OR operation is performed with the aforementioned sign value to obtain the compressed data δ, and Store the compressed data in compressed files or memory. Please refer to Figure 2. This is the process of decompressing the audio source of the present invention in PCM format (the input data is compressed data). The steps are as follows: δ 'The data can be read from the compression slot or any location in the memory. 22: Take the sign of δ and the magnitude Δ, and its formula is as follows: sign = sgn (δ), (sign = 8 & 〇x 8) A = abs (δ), (Δ = δ & 〇χ7) 23 : The indexTable updates the value of in (jex) according to its changing value, and its formula is as follows: This paper size applies the Chinese National Standard (CNS) A4 specification (210 X 297 mm). --------- Order- -------- Γ Qing first read the precautions on the back and then fill out this page) 556153 Printed by A7, Consumer Cooperatives, Intellectual Property Bureau, Ministry of Economic Affairs V. Invention Description (f) index = indexTable (δ) a = 4 24 · Calculate vpdiff 241 by continuous approximation: (delta & a) = 1 ?, if the execution result is (Yes), then _ go to 242, if the result is negative, go to 243 〇242 ·· vpdiff = vpdiff + step 243: step = step »l 244: a» l, 245: a = 0? 'If the execution result is Yes, continue to execute 25' If the execution result is negative (No), execute 241 . 25: Use the loop method to obtain the quantized signal pred 251 sr_kl = pred +/— vpdiff 252 i = 0 253 al & (〇x〇i «i) = 1? If the execution result is Yes (Yes ), Then continue to execute 254, if it is negative (No), execute 255. 254 pred = pred + sr—kl 255 sr—kl; sr—kl «li = i + l 256 i = 7 ?, if the execution result is Yes, then continue to execute 26, if it is negative (No) , Go to 253. 26: Update the value of al according to the values of b and sign, and limit al to a value between 0 and 127. The formula is: This paper size applies the Chinese National Standard (CNS) A4 specification (21〇X 297 public «)- --------- Installation -------- Order --------- (Please read the precautions on the back before filling out this page) 556153 Employee Consumer Cooperatives, Intellectual Property Bureau, Ministry of Economic Affairs Print A7 B7 V. Description of the invention (π) pred = (pred + (al »l))» 7; al = al +/- b 27: Update the value of step according to the look-up table, the formula is: step = stepTable (index ) 28: Val is the decompressed audio source and outputs it. Figure 3 is a schematic diagram of the hardware architecture of the system of the present invention, which includes the following units: A CPU (central processing unit) 31: responsible for the calculation and control of all components. A ADPCM Engirie (ADPCM engine) 32: When performing an action, first set the initial value of the ADPCM Engine 32 'the starting position of the sound file and so on. Then the CPU 31 controls the operation of the ADPCM Engine 32. When the Encode is executed, the analog to digital converter (ADC) 35 input is turned on, and the ADPCM Engine 32 is used to perform an Encode process. The compression result is stored in the surface (random access memory) 34. When decoding, ROM (read-only memory) 33 or RAM (random access 5 memory) 34 are used to generate parameter data or necessary programs. Analog to digital converter (Digital to Anai〇g

Converter ,, DAC) 36 rounds. An ADPCM / CPU synchronous execution interface: the interface between adpcm and CPU data coordination, including data bus (DataBus) and address bus (Address Bus). Please refer to Figure 4, Figure 4 is the detailed hardware architecture diagram of ADPCMEngine2 in Figure 3, which includes an arithmetic processor 32, a register 322, and a -------------- ------ Order --------- (Please read the notes on the back before filling in this page) This paper size is applicable to China National Standard (CNS) A4 (21〇X 997 " 7ν ^ γ " --------- 556153 A7

V. Description of the invention (ί /) ADPCM state 323 and one address generator 324, and its encoding method and decoding method are described as follows: The encoding method ENCODER: ADPCM State (ADPCM state) The first state of 323. The audio source is obtained from Analog to Digital Converter (ADC) 35. An arithmetic processor 321 includes an adder, a subtractor, and a shift operator, and uses the subtractor. Find the difference dif, determine the sign of the sign by dif, and then use the different processing state 321 to obtain the △ value by the continuous approximation method. The second state of ADPCM State (ADPCM state) 323: Send Δ and index to ADDRESS GENERATOR (address generator) 324, use the generated address to get the table lookup value in ROM 33, and update the register (Register ) Value of 322. The third state of ADPCM State (ADPCM state) 323: Compressed data is sent from the data bus. The first state of the decoding method DECODER, ADPCM State (ADPCM state) 323: Send the position value (address) to get the compressed data value 5. The second state of ADPCM State (ADPCM state) 323: The compressed data δ and index are sent to ADDRESS GENERATOR (Address Generator) 324, and the Chinese paper standard (CNS) A4 (210 X 297) %) -------------------- ^ --------- ^ 9. (Please read the precautions on the back before filling out this page) 556153 A7 B7 V. Description of the invention (7) The quantized value stored in ROM 33 is obtained from the address of the student, and is determined by the sign of δ. The value of vai is obtained by the adder or subtractor in the arithmetic processor 321, and val is passed through DAC 36 output. The third state of the ADPCM State (ADPCM state) 323: The data of the read-only memory ROM 33 is obtained from the compressed data δ, and the index is updated. In summary, the structural features and various embodiments of the present invention have been disclosed in detail | indicates that the present invention fully demonstrates the progress of implementation of the present invention in terms of purpose and efficacy, which has great industrial use value, and is currently According to the spirit of the Patent Law, the application of this invention has never been seen before, and the present invention fully complies with the requirements of an invention patent. The above are only the preferred embodiments of the present invention, and it should not be used to limit the scope of the present invention. That is to say, all equal changes and modifications made in accordance with the scope of the patent application of the present invention should still fall within the scope of the patent of the present invention ', I ask your reviewing committee to make a clear reference and pray for your approval.

Packing -------- Order --------- C Please read the notes on the back before filling out this page) Printed by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs

Size of this paper_㈣ 国 wr ^ NS) A4 size (2τ〇-χ 品 公 餐)

Claims (1)

55 View A8 B8 C8 D8 Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 6. Application for Patent Scope (Amendment of Patent Application Scope of Case No. 010912229) No. 1 ·-Kind A fast-adapting S-segment pulse code modulation method capable of random reading and anti-channel noise. Its applied devices include a central processing unit (QPU) ^ ADPCM engine, only 5 units of ROM (ROM), and random access Memory (test |), the method for processing compression includes the following steps: a. Inputting the sound source; b. Parameter initialization; c. Calculating the difference between the sound source X (k) and the predicted value pred (initial value is zero) dif; d · If dif is greater than or equal to zero, sign = 〇, otherwise sign = 8, then take the absolute value (vdif) of dif; e · Calculate △ by continuous approximation method; f · Use the adder to get the quantization After the signal pred; g · Update the al value (the value of the decompressed sound source data) according to the values of b and sign, and limit al to a value between 0 and 127, where b is an adjustment constant; 疋 h · According to the look-up table, Update the values of index (step value) and step (step value); and i · Data compression δ = Δ | sign, where △ is expressed as a magnitude value and the value of a sign made with the OR gate (〇r) of the logic operation, in order to achieve data compression δ. 2. The method of fast adaptive differential pulse code modulation capable of random reading and anti-channel miscellaneous gas as described in item 1 of the scope of patent application, wherein the table number in step b · is initialized to index (index value), pred (pre Set value), 厶 (large ---. 1 — II--I --------- II— lull--I c Please read the notes on the back before filling this page> This paper size applies China National Standard (CNS) A4 specification (210 X 297 ^ £ 7 A8 B8 C8 D8 Patent claim range 3 If the value of four workers is restored to the initial value (that is, I zero). N 二 专 梅 a 围The random read and anti-channel noise described in item 2 is adapted to the differential pulse code modulation method. In the initialization of step b •, another set value is set to 4, and △ and 3-human loop The differential pulse code modulation method of dependable reading and anti-channel miscellaneous gas reading speed adaptation described in item 1 of Shenyue's patent scope, wherein the step c · cutting formula is as follows: · T dif = x (k) — Pred. 5. The random read and anti-channel noise as described in Hit Liqianwei item ^ should be differential pulse code modulation method, where the method of step d · sign ^ sgn (dif); vdif = abs (dif); where The township is a sign function. When the length of f is 4 bit negative, the output value is 8, otherwise it is 0; 4 abs is to take the absolute value of the function in parentheses. The formula of the continuous approximation method includes the following: 1) Vdif>-step? (If the absolute value of the dif shirt is greater than or equal to the value), if the result is (Yes), then the result of 2) is no (No), Then skip the steps 2) Directly execute 3) ,,,, 2) Δ-te is (△ and te take or open (〇r) to judge. This paper size applies Chinese National Standard (CNS) A4 specifications (210 X 297 mm) 年 3 years η mx. AS B8 C8 D8 Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs VI. Patent application scope vdif = vdif-step (step value); · vpdiff (data gap value) = vpdiif + step; 3) step = step »l; temp = temp» l; and 4) temp = 0 ?, if the result is Yes, continue to step f ·; if the result is No, then skip to 1) Perform the continuous approximation again. 7. The fast adaptive differential pulse code modulation method that can be read randomly and resists channel noise as described in item 1 of the scope of the patent application, wherein the difference formula of step f · is as follows: 5) sr—kl = pred +/— vpdiff (data gap value); 6) i = 0; 7) al (value of decompressed audio source data) & (〇x〇l «i)? If the execution result is Yes, go to step 8), if the result is No, go to step 9) '8) pred = pred + sr__kl; 9) sr-kl = sr-kl « l; i = i + l; and 10) i = 7 ?; If the result of the execution is Yes, go to the next step-g. If the result is No, go to 7). 8. The method of fast adaptive differential pulse code modulation capable of random reading and anti-channel noise as described in item 1 of the scope of patent application, wherein the calculation formula of step g · is: pred = (pred + (al »l)) »7 ； This paper size applies to China National Standard (CNS) A4 (210 χ 297 public love) ---: ------------ installation -------- order-- ------- line (Please read the notes on the back before filling out this page) 556153 Sixth, the scope of patent application al (value of audio source data after decompression) = al +/- b. 9. The method of fast adaptive differential pulse code modulation capable of random reading and anti-channel noise as described in item 1 of the scope of patent application, wherein the calculation formula of step h · is: index = index + index Table (Δ); step (step value) = step Table (index). 10 · —A fast adaptive differential pulse code modulation method capable of random reading and anti-channel noise. The applied device includes a central processing unit (CPU), an adpcm engine, read-only memory (top of the team), and random storage. Fetch memory (rAM), the method of processing decompression includes the following steps: j · read compressed data δ; k. Fetch the sign of δ and size A; l. IndexTable updates the index according to its change value M · Calculate vpdiff by continuous approximation; η · Use the loop method to find the quantized signal pred by the adder; 〇 · Update the al (decompressed sound source data value) value according to the values of b and sign ' And limit al to a value between 0 and m; printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs; • update the value of step according to the lookup table; and q. Val is the decompressed sound source and output it. 11. The fast adaptive differential pulse code modulation method capable of random reading and anti-channel noise as described in item 10 of the scope of patent application, wherein the calculation formula of step k • is as follows: # sign ^ sgn (δ) 5 (sign = 5 &0x8); Δ = abs (δ), (Α =: δ & 0χ7). This paper size applies to China National Standard (CNS) A4 (210 χ Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 6. Scope of patent application 12. As described in item 10 of the scope of patent application, the method of fast adaptive differential pulse code modulation that can be read randomly and resists channel noise, where this step 丨 · The formula for calculating the nose is as follows: index-indexTable (Δ); a = 4 〇13 · The fast adaptive differential pulse code modulation method that can be read randomly and resists channel noise as described in item i0 of the scope of patent application, where the The calculation formula of the continuous approximation method in step m. Is as follows: 11) (delta & a) = 1 ?, if the execution result is YES, continue to execute 12), if the result is negative (N), then execute 13) 12) vpdiff (data gap value) = Vpdiff + step; 13) step (step value) = step »l: 14) a» l; and 15) a = 0 ?, if the execution result is (Yes) , Then proceed to step N · 'if it is negative (No), then execute 11). 14. The method of fast adaptive differential pulse code modulation capable of random reading and anti-channel noise, as described in item 10 of the scope of patent application, wherein the calculation method of this step η is as follows: 16) sr—kl = pred + / —Vpdiff (data gap value); 17) i = 0; 18) al (value of decompressed audio source data) & if the execution result is Yes, continue to execute 19), if it is negative (No ), Then execute 20); 19) pred = pred + sr__kl; 1 This paper size applies the Chinese National Standard (CNS) A4 specification (210 X 2971 ^ cent) '^ I: ---------- --- install -------- order --------- line (please read the notes on the back before filling out this page) The Intellectual Property Bureau of the Ministry of Economic Affairs printed the application scope of the patent for the Consumer Cooperatives 20) sr-kl = sr-kl «i; 21) 2: If the execution of the social fruit is one-time, continue to execute the right execution result is 疋, then execute step 0, if it is negative μ 15_ If the scope of patent application is tenth According to the item " Without ", the rapid adaptation of the Taolu alum, the short-span, __green, and the formula for the frequency α are: pred = (pred + (al »l))» 7 ^ al (Decompressed sound Source data value) = al +/- b ° 16 · As described in item 10 of the patent application scope, a fast adaptive differential pulse code modulation method capable of random reading and anti-channel noise, wherein the calculation formula of step ρ · is : Step (P is the distance value) = stepTable (mdex). This paper size applies to China National Standard (CNS) A4 (210 X 297 mm)