Voice pricing method and related product

Abstract

The invention provides a voice pricing method and a related product, and relates to the technical field of commodity pricing. The voice pricing method comprises the steps that preset voice information Va = {v1, v2, v3,..., vn} is acquired, and v1, v2, v3,..., vn are voice information corresponding to n pieces of commodity information; acquiring real-time voice information Vb; finding out voice information vt with the highest acoustic feature similarity with the real-time voice information Vb from preset voice information Va, and matching commodity information corresponding to the voice information vt with the real-time voice information Vb; and pricing according to a pricing rule. Related products comprise a computer readable storage medium, computer equipment, a computer program product, voice pricing equipment and a voice pricing electronic scale. According to the invention, the user can use voice to carry out pricing and settlement on the commodity according to the own language habit, so that the commodity pricing process is more efficient and convenient.

Description

Technical Field

[0001] The present invention relates to the technical field of commodity pricing, and in particular to a voice pricing method and related products. Background Art

[0002] In the process of commodity trading, pricing is required before settlement. In related devices that analyze the voice of users for pricing, generally a voice recognition system is used to process the voice issued by users, then a language model is used to convert the voice into text, and then the corresponding unit price of the commodity is matched for pricing. However, the problems in actual applications are as follows: on the one hand, the pronunciations of users with different language habits (usually so-called "local accents") may not conform to the standard, and some people usually speak unclearly or stutter, which will lead to low accuracy of voice recognition; on the other hand, people in different regions may also have different ways of referring to the same commodity, resulting in users feeling very awkward when using it. For example, for tomatoes, there are many different names in different places in China, such as "tomato", "love apple", "eggplant fruit", "red persimmon", "red", "foreign persimmon", etc. If a user usually calls "tomato" "fan shi", then when selling tomatoes, even if the voice recognition system recognizes the text "foreign persimmon", it is difficult to match the price of tomatoes for pricing. And it is very difficult for users with different language habits or pronunciation methods to use a unified name and speak standard Mandarin during transactions. The inventor believes that this is exactly the reason why the current voice recognition technology is difficult to be independently applied in transaction pricing. Although the current voice recognition pricing is theoretically feasible, it is difficult to meet the personalized pronunciation needs of different users in the actual pricing and settlement process, and often only plays some roles such as assisting users to confirm commodities, and it is difficult to completely replace manual input, barcode scanning settlement or image recognition settlement, and it is difficult to be generally accepted by users. Summary of the Invention

[0003] In the process of commodity retail trading, although the situations that the buyer and seller need to communicate may vary, the commodity information related to pricing is very limited, mainly including commodity name, commodity characteristics (such as: brand, specification model, origin, quality classification), trading volume (quantity or weight), unit, unit price, discount method, pricing instruction, etc. As long as the voice information that can distinguish the commodity information to be priced can be obtained, the traded commodity can be priced. To solve the problems in the background art, the idea of the present invention is: let the user speak the voice information of the commodity information according to his own language habit and record it in advance; during the transaction pricing, compare the real-time voice information issued by the user with the pre-recorded voice information, and find the item with the highest acoustic feature similarity in the pre-recorded voice information, then the commodity to be priced can be matched, and finally the pricing can be completed.

[0004] Therefore, the purpose of the voice pricing method and related products provided by this invention is to make the pricing process of goods more convenient and efficient.

[0005] A first aspect of the present invention is a voice-based pricing method, comprising: S1. Obtain preset voice information. a ={v1, v2, v3, ..., v n}, where v1, v2, v3, ..., v n These are the voice information corresponding to n product information items; S2, Obtain real-time voice information V b ; S3, in the preset voice information V a Find the real-time voice information V b The speech information with the highest acoustic feature similarity v t , will voice information v t Corresponding product information and real-time voice information V b Perform a match; S4. Price according to the pricing rules.

[0006] Further, in step S1, the product information is pre-recorded by the user and includes: product name, product characteristics, pricing unit, unit price, transaction volume, and pricing instruction; the preset voice information V a 为 Users pronounce product information according to their personal pronunciation habits and record it. The voice information includes the sound intensity of the voice and a waveform graph with time as the horizontal axis and amplitude as the vertical axis.

[0007] Further, in step S2, the real-time voice information V b This refers to the voice information of the product to be priced, which is sent in real time by the user during the pricing process of a commodity transaction. The voice information includes the sound intensity of the voice and a waveform diagram with time as the horizontal axis and amplitude as the vertical axis.

[0008] Furthermore, in step S3, the acoustic features can be represented by any of the following: linear predictive cepstral coefficients (LPCC), Mel frequency cepstral coefficients (MFCC), pitch period, and resonant peak.

[0009] Furthermore, in step S3, the preset voice information V a Find the real-time voice information V b The speech information with the highest acoustic feature similarity v t , will voice information v t Corresponding product information and real-time voice information V b The specific steps for matching are as follows: S31, regarding the preset voice information V aCorresponding waveform data W a ={w1, w2, w3, ..., w n}, where w1, w2, w3, ..., w n Given waveform data of speech information for n products, after framing, windowing, and noise reduction, each processed waveform is first converted into a spectrogram sequence set L. a ={l1, l2, l3, ..., l n}, where l1, l2, l3, ..., l n The spectrogram sequences corresponding to n product information items are then converted into a three-dimensional spectrogram sequence set S. a ={s1, s2, s3, ..., s n}, where s1, s2, s3, ..., s n Each of the n product information items is represented by a spectrum sequence; then, a three-dimensional spectrum sequence set S is extracted. a For each feature value in the spectrogram sequence, calculate the hash code H of the voice information feature data of the product information. a ={h1, h2, h3, ..., h n}, where h1, h2, h3, ..., h n Let H be the hash codes corresponding to n product information items. a Record and save; S32, Real-time voice information for pricing goods V b Waveform diagram W b After framing, windowing, and noise reduction, it is first converted into a spectrogram sequence L. b Then, it is converted into a three-dimensional spectrum sequence S of the product information to be priced. b Then, extract the three-dimensional spectrogram sequence S of the product information to be priced. b The feature values ​​are used to calculate the hash code H of the commodity information to be priced. b ; S33. Hash code H the information of the goods to be priced. b H, the hash code of the n already recorded product information a Calculate the similarity for each of the following: Similarity = {sim1, sim2, sim3, ..., sim...} n}, where sim1, sim2, sim3, ..., sim n These are the similarities between the information on the product to be priced and the information on the n already saved products; S34. Take sim1, sim2, sim3, ..., sim n The maximum value sim in tThis means that the pre-recorded voice information V is stored in the memory. a In the middle, voice information v t With real-time voice information V b The acoustic features have the highest similarity; real-time voice information V b With voice information v t Perform matching, voice information v t The corresponding product information is the real-time voice information V. b The information to be expressed regarding the product to be priced.

[0010] Furthermore, the specific process of step S31 is as follows: S311, Transfer waveform data W a ={w1, w2, w3, ..., w n}, where w1, w2, w3, ..., w n The waveform data of the voice information of n products are divided into frames with a set frame length of 25ms and a set frame shift of 10ms. Then, the waveform of each frame signal is multiplied by a Hamming window to obtain the windowed waveform. S312. For the windowed waveform, calculate the average noise spectrum based on a set number of frame speech signals, such as the first 5 frames. S313. Subtract the frequency of the average noise spectrum from the frequency of the vertical axis of the speech signal waveform, and set the frequency to zero when it is negative, to obtain the noise-reduced waveform as the processed waveform. S314. Convert the processed waveforms into two-dimensional spectrogram sequence sets using Fast Fourier Transform (FFT), and then plot the spectrogram sequence set L. a ={l1, l2, l3, ..., l n}, where l1, l2, l3, ..., l n Let each of the n product information items be a spectrogram sequence; then, using the corresponding sound intensity as the third dimension of the spectrogram, a three-dimensional spectrogram sequence set S is constructed. a ={s1, s2, s3, ..., s n}, where s1, s2, s3, ..., s n These are the spectrum sequences corresponding to n product information items; S315. A key point recognition algorithm, specifically the landmark algorithm, is used to extract local peak points from each 3D spectrogram as feature values. A coordinate system is determined based on the relative relationships between these local peak points. These local peak points are then represented on the coordinate system and encoded using a hash function to obtain the hash code H of the product information's voice information feature data. a ={h1, h2, h3, ..., h n}, where h1, h2, h3, ..., h n These are the hash codes corresponding to n product information items. Finally, the hash codes H are... a Record and save.

[0011] Furthermore, the specific process of step S32 is as follows: S321, Transfer real-time voice information V b Waveform diagram W b After dividing the frame into frames with a set frame length of 25ms and a set frame shift of 10ms, the waveform of each frame signal is multiplied by the Hamming window to obtain the windowed waveform. S322. Calculate the average noise spectrum based on a set quantity, such as the first 5 frames of audio signal. S323. Subtract the frequency of the average noise spectrum from the frequency of the vertical axis of the waveform graph to obtain the denoised waveform graph. When a negative frequency appears in the denoised waveform graph, set it to zero. S324, Process the waveform W b The sequences are converted into two-dimensional spectrogram sequences using Fourier Fast Transform (FFT), and then represented on a coordinate system with frequency as the horizontal axis and amplitude as the vertical axis. This coordinate system is then rotated so that the horizontal axis represents amplitude values ​​and the vertical axis represents frequency, mapping the amplitude values ​​to plot the spectrogram sequence L. b Then, using sound intensity as the third dimension, it is converted into a three-dimensional spectrogram sequence as the three-dimensional spectrogram sequence S of the product information to be priced. b； S325. Extract the three-dimensional spectrogram sequence S of the product information to be priced using the landmark algorithm. b For local peak points, determine the coordinate system based on the relative relationships between the peak points, represent the peak points on the coordinate system, and calculate the hash code H of the commodity information to be priced. b .

[0012] Furthermore, in step S34, a minimum confidence value is preset as the criterion for judging real-time voice information V. b The standard for validity; preferably, the preset minimum confidence value is 50%, that is, when sim t A match is considered successful when the accuracy rate is ≥50%, and the user's real-time voice information is considered valid pronunciation. (Real-time voice information V) b That is, voice information v t Corresponding product information; when sim t If the accuracy rate is less than 50%, it can be considered a matching failure, and the user will be prompted to repeat the pronunciation.

[0013] Furthermore, in step S4, when the user speaks real-time voice information V... bAfter matching the pre-recorded product information, the product is priced according to the product pricing rules for real-time transactions.

[0014] In a second aspect, a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the aforementioned voice pricing method.

[0015] A third aspect of the present invention is a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described voice pricing method.

[0016] A fourth aspect of the present invention is a computer program product comprising a computer program that, when executed by a processor, implements the steps of the above-described voice pricing method.

[0017] A fifth aspect of the present invention is a voice pricing device, comprising: a data acquisition module, a matching pricing module, and an output module; the matching pricing module is connected to the data acquisition module and the output module respectively.

[0018] The data acquisition module is used to collect the user's voice information and send it to the matching and pricing module; The matching and pricing module is used to execute the steps of one of the above-mentioned voice pricing methods; the matching and pricing module is also used to send the pricing result to the output module; The output module is used to output the pricing results; the output formats include digital display and / or voice broadcast.

[0019] Furthermore, the acquisition module is also used to acquire product information, which includes: product name, product characteristics, unit of measurement, unit price, transaction volume, and pricing instructions.

[0020] In most commodity transactions, users only need to state the name and quantity of the goods for pricing. However, in some transactions, weighing is required before pricing can be calculated. For example, in the trading of loose vegetables, fruits, and dried fruits, users must first weigh the goods to be priced to obtain their weight, and then calculate the price based on the unit price. Therefore, a voice-activated electronic scale capable of receiving the weight of goods for pricing is needed.

[0021] A sixth aspect of the present invention is a voice-based pricing electronic scale, comprising: a data acquisition module, a weighing module, a matching pricing module, and an output module; the matching pricing module is connected to the data acquisition module, the weighing module, and the output module respectively.

[0022] The data acquisition module is used to collect the user's voice information and send it to the matching and pricing module; The weighing module is used to measure the weight information of the goods and send it to the matching pricing module; The matching and pricing module is used to obtain the weight information of the goods through the weighing module, and then execute the steps of the above-mentioned voice pricing method; the matching and pricing module is also used to send the pricing result to the output module; The output module is used to output the pricing results; the output formats include digital display and / or voice broadcast.

[0023] Furthermore, the data collection module is also used to collect product information. Product information includes: product name, product characteristics, unit of measurement, unit price, transaction volume, and pricing instructions.

[0024] Furthermore, the matching pricing module is also used to perform the following steps: If the total weight of the goods at the Pth (P>0) weighing is x The total weight of the goods at the (P+1)th weighing is y Then the weight of the item added in the (P+1)th iteration is set to ( yx ).

[0025] The beneficial effects of this invention are: this invention does not require text parsing of the user's voice information, but only compares the voice spoken by the user during use with the voice recorded in advance to determine the product information to be priced by the user and calculate the total price of the product; different users can name the products they trade according to their own language habits, without having to follow a unified product name or pronunciation rule; users do not need to connect to the Internet during use, making it efficient and convenient. Attached Figure Description

[0026] Figure 1 This is a flowchart illustrating Embodiment 1 of the present invention; Figure 2 This is a schematic diagram of the structure of Embodiment 5 of the present invention; Figure 3 This is a schematic diagram of the structure of Embodiment Six of the present invention. Detailed Implementation

[0027] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0028] Example 1 like Figure 1 As shown, a voice-based pricing method includes the following steps: S1. Obtain preset voice information. a ={v1, v2, v3, ..., v n}, where v1, v2, v3, ..., v n These are the voice messages corresponding to n product information items.

[0029] The product information is pre-recorded by the user and includes: product name, product features, pricing unit, unit price, transaction volume, discount method, and pricing instruction; the pre-set voice information V a Record the pronunciation of product information by the user according to their personal pronunciation habits; the voice information includes the sound intensity of the voice and a waveform graph with time as the horizontal axis and amplitude as the vertical axis.

[0030] S2, Obtain real-time voice information V b .

[0031] The real-time voice information V b This refers to the voice information of the product to be priced, which is sent by the user during the pricing of the product transaction. The voice information includes the sound intensity of the voice and a waveform diagram with time as the horizontal axis and amplitude as the vertical axis.

[0032] S3, in the preset voice information V a Find the real-time voice information V b The speech information with the highest acoustic feature similarity v t , will voice information v t Corresponding product information and real-time voice information V b Perform a match.

[0033] To preset voice information V a Find the real-time voice information V b The speech information with the highest acoustic feature similarity v t First, we need to calculate the real-time voice information V. b With preset voice information V a Similarity of acoustic features; in calculating real-time speech information V b With preset voice information V a In the process of acoustic feature similarity, it is necessary to extract feature parameters of speech information. These feature parameters must be able to describe the fundamental features of the speech information as much as possible; minimize the coupling between parameter components to compress the speech information; and be simple to calculate with efficient algorithms. Linear predictive cepstral coefficients (LPCC), Mel-frequency cepstral coefficients (MFCC), pitch period, and formant peaks can all be used as feature parameters characterizing acoustic features. In this embodiment, the specific steps of S3 are as follows: S31, regarding the preset voice information V a Corresponding waveform data W a ={w1, w2, w3, ..., w n}, where w1, w2, w3, ..., w n Given waveform data of speech information for n products, after framing, windowing, and noise reduction, each processed waveform is first converted into a spectrogram sequence set L.a ={l1, l2, l3, ..., l n}, where l1, l2, l3, ..., l n The spectrogram sequences corresponding to n product information items are then converted into a three-dimensional spectrogram sequence set S. a ={s1, s2, s3, ..., s n}, where s1, s2, s3, ..., s n Each of the n product information items is represented by a spectrum sequence; then, a three-dimensional spectrum sequence set S is extracted. a For each feature value in the spectrogram sequence, calculate the hash code H of the voice information feature data of the product information. a ={h1, h2, h3, ..., h n}, where h1, h2, h3, ..., h n Let H be the hash codes corresponding to n product information items. a Recording and saving; the specific process is as follows: S311, Transfer waveform data W a ={w1, w2, w3, ..., w n}, where w1, w2, w3, ..., w n The waveform data of the voice information of n products are divided into frames with a set frame length of 25ms and a set frame shift of 10ms. Then, the waveform of each frame signal is multiplied by a Hamming window to obtain the windowed waveform. S312. For the windowed waveform, calculate the average noise spectrum based on a set number of frame speech signals, such as the first 5 frames. S313. Subtract the frequency of the average noise spectrum from the frequency of the vertical axis of the speech signal waveform, and set the frequency to zero when it is negative, to obtain the noise-reduced waveform as the processed waveform. S314. Convert the processed waveforms into two-dimensional spectrogram sequence sets using Fast Fourier Transform (FFT), and then plot the spectrogram sequence set L. a ={l1, l2, l3, ..., l n}, where l1, l2, l3, ..., l n Let each of the n product information items be a spectrogram sequence; then, using the corresponding sound intensity as the third dimension of the spectrogram, a three-dimensional spectrogram sequence set S is constructed. a ={s1, s2, s3, ..., s n}, where s1, s2, s3, ..., s n These are the spectrum sequences corresponding to n product information items; S315. A key point recognition algorithm, specifically the landmark algorithm, is used to extract local peak points from each 3D spectrogram as feature values. A coordinate system is determined based on the relative relationships between these local peak points. These local peak points are then represented on the coordinate system and encoded using a hash function to obtain the hash code H of the product information's voice information feature data. a ={h1, h2, h3, ..., h n}, where h1, h2, h3, ..., h n These are the hash codes corresponding to n product information items. Finally, the hash codes H are... a Recording is performed. In the recorded goods table, each product information Goods_Info contains several fields (field name is hash) to store the hash code value of the voice information feature data corresponding to the product information. The type is char with a length of 20, and it is initially empty.

[0034] S32, Real-time voice information for pricing goods V b Waveform diagram W b After framing, windowing, and noise reduction, it is first converted into a spectrogram sequence L. b Then, it is converted into a three-dimensional spectrum sequence S of the product information to be priced. b Then, extract the three-dimensional spectrogram sequence S of the product information to be priced. b The feature values ​​are used to calculate the hash code H of the commodity information to be priced. b The specific process is as follows: S321, Transfer real-time voice information V b Waveform diagram W b After dividing the frame into frames with a set frame length of 25ms and a set frame shift of 10ms, the waveform of each frame signal is multiplied by the Hamming window to obtain the windowed waveform. S322. Calculate the average noise spectrum based on a set quantity, such as the first 5 frames of audio signal. S323. Subtract the frequency of the average noise spectrum from the frequency of the vertical axis of the waveform graph to obtain the denoised waveform graph. When a negative frequency appears in the denoised waveform graph, set it to zero. S324, Process the waveform W b The sequences were converted into two-dimensional spectrogram sequences using Fourier Fast Transform (FFT), and then represented on a coordinate system with frequency on the horizontal axis and amplitude on the vertical axis. This coordinate system was then rotated, with the horizontal axis representing amplitude and the vertical axis representing frequency. The amplitude values ​​were mapped, and the spectrogram sequence L was plotted using quantization. b Then, using sound intensity as the third dimension, it is converted into a three-dimensional spectrogram sequence as the three-dimensional spectrogram sequence S of the product information to be priced. b ; S325. Extract the three-dimensional spectrogram sequence S of the product information to be priced using the landmark algorithm. b For local peak points, determine the coordinate system based on the relative relationships between the peak points, represent the peak points on the coordinate system, and calculate the hash code H of the commodity information to be priced. b .

[0035] S33. Hash code H the information of the goods to be priced. b H, the hash code of the n already recorded product information a Calculate the similarity for each of the following: Similarity = {sim1, sim2, sim3, ..., sim...} n}, where sim1, sim2, sim3, ..., sim n These represent the similarity between the information of the product to be priced and the information of the n already saved products.

[0036] S34. Select sim1, sim2, sim3, ..., sim n The maximum value sim in t This means that the pre-recorded voice information V is stored in the memory. a In the middle, sim t Corresponding voice information v t With real-time voice information V b The acoustic feature similarity is the highest; that is, the real-time voice information V b With voice information v t Perform matching, voice information v t The corresponding product information is the real-time voice information V. b The information to be expressed regarding the product to be priced.

[0037] It should be noted that due to various interference factors such as user pronunciation errors, for example, sometimes the real-time voice information V spoken by the user may be inaccurate. b It may be unrelated to the pricing of goods transactions, therefore a minimum confidence value needs to be preset as a criterion for judging real-time voice information V. b The standard for validity; in this embodiment, the preset minimum confidence value is 50%, when sim t A match is considered successful when the success rate is ≥50%, and the real-time voice information V is displayed. b That is, the preset voice information V a Chinese voice information v t Corresponding product information; when sim t If the accuracy rate is less than 50%, it can be considered a matching failure, and the user will be prompted to repeat the pronunciation.

[0038] S4. Price according to the pricing rules.

[0039] When the user speaks real-time voice information V b After matching the pre-recorded product information, the product is priced according to the product pricing rules for real-time transactions.

[0040] Example 2 A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of Embodiment 1 described above.

[0041] Example 3 A computer device includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of Embodiment 1 described above.

[0042] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided by this invention can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory, floppy disks, flash memory, or optical storage, etc. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory or dynamic random access memory, etc.

[0043] Example 4 A computer program product includes a computer program that, when executed by a processor, implements the steps of Embodiment 1 described above.

[0044] Example 5 like Figure 2 As shown, a voice pricing device includes: a data acquisition module, a matching pricing module, and an output module; the matching pricing module is connected to the data acquisition module and the output module respectively.

[0045] The acquisition module includes a microphone for collecting the user's voice information and sending it to the matching and pricing module; the voice information is either pre-recorded voice information stored by the user or real-time voice information emitted during transaction pricing; the acquisition module is also used to collect product information through computer equipment, the product information including: product name, product characteristics, pricing unit, unit price, transaction volume, and pricing instructions.

[0046] The matching and pricing module includes a memory and a processor; the matching and pricing module is used to execute the steps of the above embodiment one; the matching and pricing module is also used to send the pricing result to the output module.

[0047] The output module includes a speaker and a display for outputting pricing results; the results are output in the form of digital display and voice broadcast.

[0048] Example 6 like Figure 3 As shown, a voice-based pricing electronic scale includes: a data acquisition module, a weighing module, a matching pricing module, and an output module; the matching pricing module is connected to the data acquisition module, the weighing module, and the output module respectively.

[0049] The acquisition module includes a microphone for collecting the user's voice information and sending it to the matching and pricing module; the voice information is either pre-recorded voice information stored by the user or real-time voice information emitted during transaction pricing; the acquisition module is also used to collect product information through computer equipment, the product information including: product name, product characteristics, pricing unit, unit price, transaction volume, and pricing instructions.

[0050] The weighing module includes a weight sensor for measuring the weight information of the goods and sending it to the matching pricing module; The matching and pricing module includes a memory and a processor; the matching and pricing module is used to obtain the weight information of the goods through the weighing module and execute the steps of Embodiment 1 above; the matching and pricing module is also used to send the pricing result to the output module.

[0051] The output module includes a speaker and a display for outputting pricing results; the results are output in the form of digital display and voice broadcast.

[0052] The method of using the voice-based pricing electronic scale in Embodiment Six is ​​as follows: Before use, the user connects the electronic scale to a computer device, saves the product information to the matching pricing module through the computer device, and then the acquisition module pronounces and records the information of each product one by one, that is, records the preset voice information to the matching pricing module. During the weighing and pricing process, the user puts the product into the weighing module and then speaks the real-time voice information of the product to be priced into the acquisition module. The matching pricing module compares and matches the real-time voice information with the recorded preset voice information, calculates the price based on the unit price of the product and the weight data measured by the weighing module, and finally the output module announces the pricing result. When recording the preset voice information, the user can pronounce and name the product information according to their own pronunciation habits. For example, when recording the voice information of the "product name" of tomatoes, the user's pronunciation can be any one or more of the following: "xi hong shi", "fan shi", "fanqie", "hong", etc. Suppose the user records the voice information of the product name of tomatoes as "hu die". Then, during the weighing and pricing process, if the real-time voice information obtained by the pricing module is "hu die", the product information of tomatoes can be matched for them. Then, the unit price of tomatoes can be retrieved, and the transaction price of the tomatoes to be priced can be calculated based on the weight of the tomatoes measured by the weighing module.

[0053] It should be noted that when using traditional voice recognition pricing devices, the real-time voice information spoken by the user is the product name, and the device matches the product unit price after recognizing the product name. However, the product information described in this embodiment six includes: product name, product characteristics, pricing unit, unit price, transaction volume, discount method, pricing instruction, and the real-time voice information (V) spoken by the user. b It can be the product name, unit price, or other information, which provides users with more convenience. For example, when selling tomatoes, a user can directly state the unit price of the tomatoes without saying the product name. Similarly, the voice-based pricing electronic scale in Embodiment Six can directly calculate the price of the product (i.e., tomatoes) in the transaction using the unit price and the weight information provided by the weighing module.

[0054] In some scattered transactions, users sometimes combine the pricing of many different types, weights, and prices of goods. For example, a customer might buy a bunch of scallions, two radishes, and a block of tofu. The usual practice is to first weigh the scallions to get their price, then remove the scallions, add the radishes to get their price, then remove the radishes, add the tofu to get their price, and finally sum the three prices to get the total price. This is very inconvenient for the user. Therefore, in the voice-based pricing electronic scale of this embodiment six, the matching pricing module is also used to perform the following steps: If the weight of the goods is at the Pth (P>0) weighing time... x The total weight of the goods at the (P+1)th weighing is y Then the weight of the item added in the (P+1)th iteration is set to ( yx ).

[0055] The above steps enable the voice-based pricing electronic scale of Embodiment Six to simultaneously weigh and price multiple goods. For example, a user can first put scallions into the weighing module and say "scallions" into the data acquisition module, then put in radishes and say "radishes," then put in tofu and say "tofu." Each item does not need to be removed immediately after being put into the weighing module. Instead, the user can remove all items at once and deliver them to the customer after all items have been weighed and priced. This makes the user's operation process more convenient and efficient, which is unmatched by other types of electronic scales that rely on image recognition.

[0056] The above embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

Claims

1. A voice-based pricing method, characterized in that, include: Get Preset Voice Message V a ={v1, v2, v3, ..., v n }, where v1, v2, v3, ..., v n These are the voice information corresponding to n product information items; Get real-time voice information V b ; Preset voice information V a Find the real-time voice information V b The speech information with the highest acoustic feature similarity v t , will voice information v t Corresponding product information and real-time voice information V b Perform a match; Pricing shall be conducted in accordance with the pricing rules.

2. The voice-based pricing method according to claim 1, characterized in that: The acquisition of preset voice information V a ={v1, v2, v3, ..., v n }, where v1, v2, v3, ..., v n Among the voice messages corresponding to n product information items, the product information is pre-recorded by the user, and the pre-set voice message V... a Users record and save the pronunciation of product information according to their personal pronunciation habits. The voice information includes the sound intensity of the voice and a waveform graph with time as the horizontal axis and amplitude as the vertical axis.

3. The voice pricing method according to claim 1, characterized in that: The acquisition of real-time voice information V b In the middle, real-time voice information V b 为 The voice information of the product to be priced, sent by the user in real time during the pricing process of the commodity transaction, includes the sound intensity of the voice and a waveform graph with time as the horizontal axis and amplitude as the vertical axis.

4. The voice pricing method according to claim 1, characterized in that: The preset voice information V a Find the real-time voice information V b The speech information with the highest acoustic feature similarity v t , will voice information v t Corresponding product information and real-time voice information V b To perform matching, including: For preset voice information V a Corresponding waveform data W a ={w1, w2, w3, ..., w n }, where w1, w2, w3, ..., w n Given waveform data of audio information for n products, after framing, windowing, and noise reduction, each processed waveform is first converted into a spectrogram sequence set L. a ={l1, l2, l3, ..., l n }, where l1, l2, l3, ..., l n The spectrogram sequences corresponding to n product information items are then converted into a three-dimensional spectrogram sequence set S. a ={s1, s2, s3, ..., s n }, where s1, s2, s3, ..., s n Each of the n product information items is represented by a spectrum sequence; then, a three-dimensional spectrum sequence set S is extracted. a For each feature value in the spectrogram sequence, calculate the hash code H of the voice information feature data of the product information. a ={h1, h2, h3, ..., h n }, where h1, h2, h3, ..., h n Let H be the hash codes corresponding to n product information items. a Record and save; Real-time voice information for pricing goods V b Waveform diagram W b After framing, windowing, and noise reduction, it is first converted into a spectrogram sequence L. b Then, it is converted into a three-dimensional spectrum sequence S of the product information to be priced. b Then, extract the three-dimensional spectrogram sequence S of the product information to be priced. b The feature values ​​are used to calculate the hash code H of the commodity information to be priced. b ; Hash code H for the information of the goods to be priced b H, the hash code of the n already recorded product information a Calculate the similarity for each of the following: Similarity = {sim1, sim2, sim3, ..., sim...} n }, where sim1, sim2, sim3, ..., sim n These are the similarities between the information on the product to be priced and the information on the n already saved products; Compare sim1, sim2, sim3, ..., sim n The maximum value sim in t , will voice information v t The corresponding product information is used as the product information to be priced.

5. A computer-readable storage medium, characterized in that, A computer program is stored on a computer-readable storage medium, which, when executed by a processor, implements the steps of the method according to any one of claims 1-4.

6. A computer device, characterized in that: It includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of the method according to any one of claims 1-4.

7. A computer program product, comprising a computer program, characterized in that: When the computer program is executed by a processor, it implements the steps of the method described in any one of claims 1-4.

8. A voice-based pricing device, characterized in that, include: The system comprises a data acquisition module, a matching and pricing module, and an output module; the matching and pricing module is connected to the data acquisition module and the output module respectively; the data acquisition module is used to acquire voice information; the matching and pricing module is used to execute the steps of the method according to any one of claims 1-4; and the output module is used to output the pricing result.

9. A voice-activated electronic scale, characterized in that, include: The system comprises a data acquisition module, a weighing module, a matching and pricing module, and an output module; the matching and pricing module is connected to the data acquisition module, the weighing module, and the output module respectively; the data acquisition module is used to acquire voice information; the weighing module is used to measure the weight information of the goods; the matching and pricing module is used to obtain the weight information of the goods through the weighing module and execute the steps of the method described in any one of claims 1-4; the output module is used to output the pricing result.

10. A voice-based pricing electronic scale according to claim 9, characterized in that, The matching and pricing module is also used to perform the following steps: If the total weight of the goods at the Pth (P>0) weighing is x The total weight of the goods at the (P+1)th weighing is y Then the weight of the item added in the (P+1)th time is set to ( yx ).

Images