Patent Grant
10817918
The current invention relates to a system for creating custom fragrances, and specifically to a system for creating custom fragrances in an objective and repeatable process.
The vast majority of fragrances are made by specific formulas that are mass marketed to the public. These fragrances may be made into perfumes and colognes, lotions, soaps, body washes and other personal hygiene products. These are usually determined by the companies and not custom made by the customer.
There have been some attempts to allow a user to make custom products such as U.S. Pat. No. 6,402,120 B1 Swaab, Jun. 11, 2002, “Apparatus for Blending and Fabricating Personalized Lipstick”. This deals with custom color for lipstick and does not deal with the creation of custom fragrances.
US Patent Application 2008/0027820 A1 Brill, Jan. 31, 2008, “Methods and Systems for the Creation of Consumer-Customization and Commercialization of Goods” discloses customization of wines but does not disclose any description of how to customize fragrances.
US Patent Application 2014/0032356 A1 Chuang et al., Jan. 30, 2014, “Method and System for Selling Custom Products” deals with the general idea of customization of products, but requires the customer to send their customization preferences to a central server. Obviously, there is no way to electronically sample fragrances, so this cannot apply to the customization of fragrances.
U.S. Pat. No. 6,516,245 B1 Dirksing et al., Feb. 4, 2003, “Method for Providing Personalized Cosmetics” and US Patent Application 2003/0014324 A1 Donovan et al., Jan. 16, 2003, “Techniques for Synthesizing and Distributing Personal Care Products” have a similar problem as Chuang above, being that there is no way of remotely sampling scents.
U.S. Pat. No. 5,031,764 Meador et al., Jul. 16, 1991, “Apparatus for Designing Personalized Perfume” discloses a group of blotters held together as a single unit for sampling scents. However, blotters are not intended to be added or deleted and therefore can only be used as a fixed group of scents.
US Patent Application 2008/0131858 A1 Gordon, Jun. 5, 2008, “Method and Apparatus for Creating a Custom Blended Fragrance” describes a method of creating custom fragrances based upon a personality test. This tends to be very subjective and not accurate.
U.S. Pat. No. 5,626,155 Saute, May 6, 1997, “Method of Creating Fragrances In Situ” requires that the mixed scents be applied on the customer's skin. This will only apply to fragrances mixed in-situ.
These attempts to create perfumes are typically subjective and random. A user would simply smell various scents and mix them without regard to groupings or volatility of the scent. In which case, some fragrances would dominate the others or the end result of the fragrance composition would be poorly structured thus not having an appealing odor.
There was also little consideration given to the time-changing aspects of the fragrance. Therefore, a fragrance which is initially appealing may become less appealing over the course of the day.
Also, there has been little effort made to predict and reduce the amount of conflicting fragrances.
When customers buy premixed fragrances that are not custom made, they do not have a choice about their contents. They cannot adjust the fragrances to be more gender specific or gender neutral (i.e. unisex), adjust the tenacity, or concentrations. The customers also do not have the ability to create new fragrances for special occasions, seasons or holidays.
Currently, there is a need for a more accurate system for creating custom fragrances which is less subjective, takes into account the customer's desires, the time-changing aspects of the fragrance, and can save the parameters of each customer's unique scents.
In one aspect, the invention relates to a method of creating a personalized custom fragrance for a customer by providing a ‘Perfumer's Organ’ of variable strength scent liquids and normal strength scent liquids, grouped by family and by volatility category;
allowing the customer to select and place scent liquid on a blotter to smell the scents for a plurality of scent liquids;
having the customer select at least two scents from at least two different volatility categories that the customer would like to be in the custom fragrance;
allowing the customer to smell the mixture of all of the selected scents simultaneously;
allowing the customer to delete scents that they would no longer like to be part of custom fragrance;
allowing the customer to add scents to be part of the custom fragrance. It also includes again providing the currently selected scents simultaneously to the customer to allow the customer to smell the mixture of all of the selected scents;
determining an amount of each selected variable potency scent liquid to use based upon a prestored scent table;
determining the total number of drops N to be used for the selected container;
determining the number of drops of all of the variable potency scent liquids to be added to the selected container;
subtracting the number of drops of the variable potency scent liquids intended to be used from the total number of drops for the selected container N to result in the remaining drops (Nremain);
determining the number of normal potency scents selected (n) and dividing the remaining drops (Nremain) by the number of normal potency scents selected (n) to result in the number of drops for each of the n normal potency scent liquids to add to the container; and mixing the determined amounts of each selected scent liquids with a filler material to result in the custom fragrance.
The current invention may also be described as a device for interacting with an employee and a customer to create custom fragrances having an input device for receiving scents selected by the customer;
a memory having a pre-stored “training data” being a plurality of equations of desired combinations of variable potency scents and normal potency scents each multiplied by scaling coefficients (αi) and compatibility coefficients (βi); and
a CPU coupled to the input device and the memory. The CPU having an internal executable program causing it to:
The invention may also be describe as a method of mixing fragrances from variable potency scents and normal potency scents selected by a customer comprising the steps of: acquiring a scent table of variable potency scents indicating the volumes “Xi” of each variable potency scent “i” to use for several different sized containers to be used in creating the custom fragrance;
for each scent “i” where i=1 through n variable potency scents, looking up a volume Xi in the scent table;
determine the total variable potency volume XT for all selected scents;
adjusting the volume of each scent i by
S*(Xi/XT)*2.67=Ni
where Ni is the number of drops for each variable potency scent “i”, S=formula size (mL);
determining the remaining amount Nremain to be filled with the normal potency scents;
acquiring a training set of general equations for highly desirable fragrances each indicating the relative amounts Ni of each scent to use, wherein the equations have a volume coefficient and a compatibility coefficient for each of the scents;
fitting a generalized equation N*Y*Σ(αi*βi)=N, to the training set of equations to determine the volume coefficients αi and the compatibility coefficients βi, where N=total drops in formula, Y=filling factor, αi=scaling coefficient for each scent in the formula, and βi=compatibility coefficient for each normal potency scent in the formula;
using the equation N*Y*αi*βi=Ni to identify the number of drops of each normal potency scent to use in completing the custom fragrance.
It will be appreciated that the illustrated boundaries of elements in the drawings represent only one example of the boundaries. One skilled in the art will appreciate that a single element may be designed as multiple elements or that multiple elements may be designed as a single element. An element shown as an internal feature may be implemented as an external feature and vice versa.
Further, in the accompanying drawings and description that follow, like parts are indicated throughout the drawings and description with the same reference numerals, respectively. The figures may not be drawn to scale and the proportions of certain parts have been exaggerated for convenience of illustration.
Theory
In the current process, the customer decides the parameters of the fragrances.
For example, men typically gravitate to the scent families of earthy/woody, herbal, fresh and clean and citrus.
For tenacity, the customer 5 can decide how long the fragrance will remain by selecting higher or lower notes (higher and lower volatility scents). The fragrance concentration level can also be adjusted from EDP (eau de parfum) to EDT (eau de toilette).
Fragrances can also be created for a certain occasion such as a wedding day, date night, etc.
System & Method
These scents may have normal potency or variable potency. The normal potency scents each have the same scent intensity when the same amount is used. (Larry—Also though, when a custom fragrance is blended, the potency of the individual scent may change depending on the levels used and/or the other scents in the formula.
However, the variable potency scents require varying amounts to produce the same scent intensity. These are experimentally determined and tabulated. These tables are pre-stored with other information in the initial information 7 of
The customer 5 interacts with the Perfumer's Organ 110 and the employee 3 to select various scents to be used to create a custom mixture of scents, also referred to as a custom fragrance.
The customer 5 determines the scents to be used in the mixture, the type of product, the container size and if a perfume or cologne or a personal care product, and the concentration level of aromatic compounds (eau de parfum, eau de toilette). The customer 5 provides this information to the employee 3.
In a first embodiment, the employee, referred to as a “Scent Designer” 3 determines which scents selected by the customer 5 are normal potency scents and which are variable potency scents. The employee 3 reads from a predefined table in a training manual referred to in
The Scent Designer asks the customer their likes and dislikes, what scents they want to highlight or make any scents more pronounced than others (i.e. the customer 5 would like more lilac or want lilac to be the overall aroma of the fragrance). The opposite is also true in that if a customer wants to tone down a scent, only wants to smell a little bit of it, that can be done as well. The Scent Designer will incrementally increase or decrease the number of drops according to the customer's request.
Scent Designers 3 may adjust the formula based on how different scents react or perform with each other. For example, a customer 5 may say they want a large amount of patchouli in their formula. Patchouli is recommended at 8 drops in the 30 ml bottle. Since its strength is so overpowering of the other notes, the amount used must be within a predetermined range of the amount indicated in the scent table. For example, if the predetermined range is 10%, the maximum amount used would be no more than 110% of the amount in the scent table. For most cases, the predetermined range is 5%-15%.
This resulting number is evenly divided by the number “n” of normal potency scents. This is the number of drops “Ni” of each of the normal potency scents to be used.
A filler material which corresponds to the product selected by the customer is added to the container 120 leaving room for adjustment. For perfume and cologne products, the filler material is typically denatured alcohol.
In another embodiment, the employee 3 interacts with a computing device 140 to enter the customer's selections. The computing device 140, in this embodiment has an input-output device, such as a screen 141 for receiving input from the employee and for displaying information to the employee. It communicates information with a CPU 143. The CPU interacts with an internal memory 145 which may have a stored database, and optionally with a remote storage device 150 which also may have a stored database.
The CPU 143 employs the input provided by the employee 3 and initial information 7 from the training manual and stores them in the database(s). The CPU 143 determines the amount of each of the scents to be used in the custom fragrance.
In another embodiment, the employee 3 places the amounts of each of the scents indicated by the computing device 140 into the container 120 specified by the customer 5.
The employee 3 then also places the amount of the filler material in the container 120.
At this point the container is mixed and provided to the customer 5. The customer 5 then is allowed to provide feedback on which scents to increase or decrease. The employee 3 then adds an additional amount of the scent indicated to be increased. In the case where the customer 5 would like to decrease a scent in the mixture, then the employee 5 increases the amount of all other scents except the one that is intended to be decreased.
The customer 5 is allowed to smell the mixture and repeat the previous adjustment steps until they are satisfied.
The filler material 135 is then added to fill the container to result in the custom fragrance 170 which is the end product.
In an alternative embodiment, an automated filling device 160 is controlled by the computing device 140. The automated filling device 160 receives the container 120 selected by the customer 5, the filler material 135 corresponding to the selected product to be made and the calculated amount of each scent to be added to the container 120. This automatically prepares the custom fragrance interactively created by customer 5.
In an alternative embodiment, the current mixture is provided to the customer 5 to smell. The customer 5 indicates scents to increase and those to decrease. These are provided to the computing device 140 by the employee 3. The container 120 is inserted back into the automated filling device 160 and the computing device 140 adds scents to adjust the mixture according to the instructions from the customer 5.
The embodiment of the Perfumer's Organ 110 shown in
However, since the top note 301 has the highest volatility, it quickly dissipates and loses the most of the original amount. Similarly, the middle note 303 has a reduced amount as compared with the base note 305. Therefore, at time t1, the top note 301 is at about 5% of its original amount, the middle note 303 is at about 25% of its original amount and the base note 305 is at about 50% of its original amount. Therefore, at time t1, the dominant scent is the base note 305. Middle note 303 contributes only half of the observable fragrance and the top note 301 is almost indistinguishable.
At time t2, the fragrance perceived is almost entirely attributed to the bottom note 305. Therefore, by quantifying the strengths of the scents over time, the customer can select what a fragrance will be initially and what the fragrance will be at later times during the day. It is possible to select scents and notes to have a desired mixture, for example, at lunch time, or for a dinner party. Therefore, the employee educates the customer as to the time-changing nature of different notes and the customer takes this into account when selecting the scents and notes to be used.
In another embodiment, volatility curves, such as those shown in
The process of creating custom fragrances can be embodied in the steps of the flowchart of
The Employee Explains the Process
In order to create custom fragrances, there is some interaction between the employee and the customer. First the employee meets the customer and escorts the customer to a table which has the Perfumer's Organ.
In
The employee then indicates that when building a fragrance, it is best to have some top notes, some middle notes and some base notes in your blend. This helps to create a balanced and well-structured scent.
Next, the employee will show the customer the fragrance blotters that will be used to sample the scents and later blend the notes together. The blotters have a portion in which information may be written, such as the scent name. There is a portion which can be handled and a dipping portion which is dipped into a scent.
The Customer Chooses Initial/Subsequent Scents
In step 440, the customer would dip the blotter into a scent and smell the scent. If (s)he likes the scent, an identifying name is written on the blotter and the blotter is selected. The customer continues to select preferably 3 to 7 scents and preferably at least one scent for each note in step 450.
In step 460, all selected scents/blotters are attached together with the clip with the blotters slightly fanning out. The clip and blotters are waved through the air to mix the scents.
Sample with Clip and Blotters
The customer is then instructed to smell the mixed scents in step 470.
In step 480, the customer is allowed to adjust the scents chosen. This can be done by removing a blotter from the clip or adding another blotter with a different scent to the clip.
The customer then smells the adjusted scents in step 485.
Customer Adjusts Scent Intensities
In optional step 490, the customer can adjust the scent intensity. The scent intensity can be increased by adding blotters of the scent to be increased, selecting blotters of more surface area, or using blotters with higher absorbency to absorb more scent liquid, using marked blotters and dipping the blotters farther into the scent liquids, as indicated by the markings to result in more scent liquid on the blotter.
The scent intensity can be decreased by removing one or more blotters having the same scent, selecting blotters having less surface area, or using blotters with lower absorbency to absorb less scent liquid, or using marked blotters and dipping the blotters less of the distance into the scent liquids, as indicated by the markings to result in less scent liquid on the blotter.
Another way of reducing the strength of a scent would be to cut off a portion of the dipping end of the blotter having the scent oil.
The strength of a scent may also be reduced by limiting its interaction with the airstream as it is being waved. This may include folding the dipping portion of the blotter back upon itself, reducing the airflow over the dipping portion of the blotter, or shortening the blotter such that it spans a smaller arc when waved.
In step 495, the customer smells the blotters held by the clip and decides if they are done adding or deleting scents in step 500. If they are done, processing continues at step 510 of
In
Scents Chosen by the Customer are Used to Create the Fragrance
The customer indicates the scents to use and the size of container and the concentration of the fragrance, if a perfume or cologne that (s)he desires. The fragrances may then be manufactured.
Manufacture of the Custom Fragrance Using a Table-Based Method
In step 520, the scents are identified as either variable potency scents or normal potency scents.
Adding Variable Potency Scents
The variable potency scents provided in an equal amount do not provide equal impact upon the mixture. Therefore, there must be a table or equation indicating the amounts to use of each variable potency scent. The scents used in the Perfumer's Organ are listed below by name, classification, Family, color code in the Perfumer's Organ, Gender, Seasonality and Potency. The variable potency scents are indicated in the column marked “Potency”:
Patchouli
Vanilla
Gardenia
Eucalyptus
The amounts of the scents may be measured in terms of drops. “N” is the total number of drops of all scents in a container. “Ni” is the number of drops for a normal potency scent. “Xi” is the number of drops for variable potency scent “i”. Therefore, in step 530 for an equal impact, or perceived fragrance strength, the number of drops for a given container size is found in the table and used. This is repeated for all variable potency scents. The total number of drops for the variable potency scents for “n” scents is X=sum(Xi) from i=1-n.
Adding Normal Potency Scents
The remaining number (Nremain) of normal potency scents drops to be added to a container of this size is Nremain=N−Xt drops. These are added in equal proportions in step 540. Therefore, if Nremain=100 drops and there are 5 remaining normal potency scents, there would be 20 drops of each of the five remaining normal potency scents.
The amounts determined for each of the variable potency scents and the normal potency scents are added to the selected container in step 550.
Adding Filler Material
A carrier material is added to the variable potency scent drops and the normal potency scent drops in the container. For a perfume or cologne the carrier material may be an alcohol solution.
The overall perceived strength or concentration can be adjusted as per the request of the customer. Therefore, different entries will be listed in the initial information (7 of
The concentration can also be adjusted based on the age of the customer. Customers aged 12 and under receive the eau de toilette concentration. The typical concentration for adults is an eau de parfum concentration level of 15-20%. In other embodiments, other concentration levels may also be offered.
For lotions, creams or other body care products the filler material may be a topical oil base, or may be a wax-based substance for other products. The container is then filled to about 80% of its capacity in step 560, leaving room for adjustment.
Customer Samples the Mixture
The customer is then allowed to smell the mixture in step 570 and provides feedback in step 580.
Adjusting Upward or Downward
In
If the customer would like to have a certain scent diminished, then a few drops of all other scents in the mixture will be added in step 600.
In step 610 the customer smells the adjusted mixture.
In step 620 the customer determines if the adjustment is finished. If so, processing continues at step 630.
If further adjustments are required, then steps 580 through 620 are repeated until the customer is satisfied with the mixture. Once the adjustments are completed, the information is recorded with a unique ID number. Every customer receives a unique ID number for every formula created. This is done for identification and categorization of formulas and reordering purposes. The customer receives a key tag with their ID number on it to simplify the reordering process. This information may also be stored in the computing device 140 and used for various purposes and analyses. For example, the most popular scents may be determined, the most common combinations, etc. This may also be used for inventory and restocking of scent materials. Finally, these may be used for advertising, marketing and consumer research purposes.
In step 630, the mixture is topped off with the filler material.
Since the customer has created the personalized fragrance, they may name the fragrance. In step 640, personalized scent name labels are printed with the name the customer has chosen and the personalized labels are affixed to the finished product. This adds another level of customization and personalization. The process is completed in step 650.
In step 543, the number “n” of normal potency scents is determined.
In step 545, the amount for all of the normal potency scents “Nremain” is divided equally for the “n” normal potency scents in the mixture.
In step 536, the amount for a standard volume of each scent Xi to use is taken from the table.
In step 537 all of the “Xi” are summed to get the total XT.
In step 538, the ratio of “Xi/XT” is adjusted for the container size “S” and a constant to result in the number of drops of each variable potency scent “Ni”.
Formula Prediction Algorithm
In an alternative embodiment, the manufacturing of a pleasant fragrance incorporates the following computations.
All of the scents are identified by an identifier number “i”. A number of formulas or mixtures which have had significant success are selected and are referred to as the “training set” or the “training data”. Each mixture is a combination of the amounts of at least two of these scents. (It is possible for the amount of one or more scents in the training data to be zero or near zero.)
The intent is to take a generalized equation (1) and to iteratively adjust the coefficients αi and βi to fit it to the training data.
Modeling of Perfume Formulas
The composition of any given perfume formula is modeled by the following equation:
N*Y*[(α1*1)+(α2*β2)+(α3*β3) . . . +(αn*βn)]=N Eq. (1)
for the n scent oils chosen by the customer. Equation 1 may also be written in another form as:
N*YΣ(αi*βi)=N Equation 1. General Equation
N=total drops in formula
Y=filling factor
αj=scaling coefficient for each scent in formula
βi=compatibility coefficient for each scent in formula
The filling factor “Y” is the percentage that the chosen container is filled, allowing extra room for additional adjustments.
There are a number of known algorithms that fit a generalized equation to a set of training data (or training equations). Stochastic algorithms are uniquely suited for this problem because they can search a very large solution space with no prior domain knowledge. Preferably, a Stochastic algorithm is used which will introduce small, random changes in the variables being searched and keep those variable values that are beneficial.
The process starts with initializing αi and βi to “1”. Plugging these into Equation (2) results in a predicted set of the number of drops of each scent (Ni). The predicted set is then compared to the number of drops for all mixtures in the training set. The difference between the number of drops for the predicted formula and the first training mixture is calculated. This is repeated for all scents for all formulas in the training set. The total error is accumulated and the average of all of these errors is calculated.
Next, αi is changed slightly and the process repeated. This is done iteratively until the change in error is below a predefined threshold. This indicates that the equation is converging to an answer.
Once this point is met, then the other coefficient βi is changed and the process repeated for coefficient βi. This is done iteratively until the change in error is below a predefined threshold, indicating that it is converging to an answer.
At this point, αi and βi are now defined and can be used in Equation (1).
Since the result is expressed in drops of each scent, it must be normalized for various sized containers using N, the total number of drops in a container and “Y” a filling factor.
N, the total drops, is determined by the formula size in milliliters. The standard values at each size can be found in the training manual which is the same as the initial information (7 of
N, αi, and βi are now known, so the only free variable is Y, the filling factor. The filling factor guarantees that the sum of drops for individual scents is equal to the desired total drops, regardless of the number of ingredients in the formula.
General Equation (1) can be converted into Equation (2) which is specific for each scent.
N*Y*αi*βi=Ni Equation 2: Drops for each scent
Ni=drops for each scent
Input
Below is a list of input provided to the program.
Using the inputs above and Equation (2), a new mixture is produced using the inputs from the user listed above to produce a desirable mixture similar to the list of scents provided by the user.
Derivation
The derivation of this Eq. (1) is clarified by considering a formula where each scent is distributed in constant proportion: the equation would be N*Σ(Xi)=N, where Xi is the proportion for each scent. To get Eq. (1), we make the proportion the product of two coefficients (α□, β□) and the free variable Y The filling factor Y allows us to guarantee that the formula contains the desired number of total drops and lets us specify an arbitrary range for alpha and beta.
Both alpha (αi) and beta (βi) are the coefficients and are calculated by the optimization algorithm. αi is a scaling coefficient for each scent that theoretically represents relative strength. βi is a compatibility coefficient that represents how well a given scent scales with the other scents in the formula. βi is the average of the beta values for the other scents in the formula. For example, in a formula with 5 scents the βi for each scent would be the average of the beta values for the other 4 scents.
It should be noted that the compatibility coefficient of scent A with scent B (for any two scents A and B) is not the same value as the compatibility of scent B with scent A. In other words, each scent possesses as many compatibility coefficients as there are other scents in the system.
S*(Xi/XT)*2.67=Ni and S*2.67=N Equation 3: For Variable potency scents only
S=formula size (mL)
Xi=drops per mL for a given scent
XT=the sum of Xi's in the formula
Ni=drops for scent i
Equation 3 is for variable potency scents only. These are scents that typically dominate over other scents, so their amounts (drops) used must be adjusted. When a formula contains only variable potency scents, the number of drops for each scent is scaled up from the recommended values so that the total drop count is not extremely low. Since there are approximately 2.67 drops per ml, the total number of drops for a formula that contains only variable potency scents is simply size in mL*2.67, where 2.67 is a constant specified by the client. The Xi values are derived from a variable potency scent table which indicates the relative amounts of scents to be used based upon their inherent odor strengths. Tables of the initial information (7 of
Summary of Automated Procedure
Below are the steps of one procedure for creating custom fragrances from the customer scent selections, and container size according to one embodiment of the present invention.
1.) Calculate the total drops (N) for the formula size with the following equation:
Total drops=formula size (mL)*number of drops/mL
This results in a filling factor of 87.5% of the recommended size, leaving room for adjustment.
2.) Calculate the number of drops for each variable potency scent “i” according to Eq. (3). As in the previous step, the number of drops for each variable potency scent is the product of the formula size in milliliters and the drops per milliliter for that scent (from the variable potency scent table).
3.) Calculate the in-formula compatibility value for each scent (beta) as the average of that scent's compatibility factors with the other scents in the formula.
βi=βk/(# of scents in formula−1) Eq. (4)
4.) Adjust current scaling factors (αi) by the value of the compatibility factor (βi) to result in a new scaling factor (α).
new α=current α*β Eq. (5)
5.) Use the equation in Section “5b” below for formulas with variable potency scents, and the equation in section “6” below for formulas with normal potency scents.
S=formula size (mL)
Xi=drops per mL for a given scent
XT=the sum of Xi's in the formula
Ni=drops for scent i
6.) Calculate the drops for normal potency scents based on the Equation (2).
N*Y*αi*βi=Ni
Ni=Drops for each scent
“Y” is defined in Eq. (6), αi is defined in Eq. (5) and βi is defined in Eq. (4).
Add all scents to formula and return completed formula.
6b. Empirical Method of Manufacture
6b 1. Acquire Input Data
In
A listing of scents and amounts of the scents that have been successful in the past, referred to as “training data” is also input to the computing device, in the same step.
The desired container size is identified and the type of product desired. Optionally, the type of product may be a perfume, cologne, a lotion, soap, etc. may also be input in this step.
In step 1003, a general equation Eq. (1) is initiated with alphas and betas=“1”.
In step 1005 the predicted amount of each scent is determined from general Eq. (1) for all scents.
An error is calculated in step 1007 between the predicted amount and the training data amount for the same scent. This error is summed for all equations of the training data and is used to determine a total error, and an average error that are stored in step 1009.
The alpha coefficients (αi) are slightly changed in step 1011 and the error is calculated as indicated for step 1007 in step 1013. If the average error is lower than the previously smallest average error to this point, it is stored with the coefficients with αi in step 1015.
In step 1017, if the change in average error for the current iteration is less than that of the previous iteration, then the answer is converging and processing then continues at step 1019 of
If the error in step 1017 is not less than the predetermined threshold, there should be more iterations and processing from steps 1011 through 1017 are repeated.
Repeat for Beta Coefficients
The process continues at step 1019 of
In step 1021, βi is calculated.
The Filling Factor is determined.
In step 1023, the filling factor “Y” can now be determined, since “N”, “αi” and “βi” are now known.
Determining the Amounts of Scents
Plugging in the αi, βi, and user scent choices into Eq. (3) results in amounts of each scent to put into the selected container. Finally, the drops “Ni” of the normal potency scents “i” can be determined in step 1025.
The ‘hill climbing algorithm’ is one algorithm that may be used to determine these alpha and beta coefficients for the general equation. There are other commonly known algorithms which also may be used to fit the generalized Eq. (1) to a set of training data.
Software Implementation
The software implementation of a portion of the system determines perfume formula composition. In addition, it acts as a store management tool. The system algorithmically optimizes a set of formula composition variables based on previous formulas. Information about formulas, orders, customers, and employees is stored in a database (currently implemented in MySQL), which can be edited through an interface provided to administrative users.
The system is designed to predict how to create a fragrance (mixture of scents) based on previous ones, even for combinations of ingredients that were never used. This “prediction engine” must work in the context of a larger application which will have additional functions.
Overview
Referring to
The first stage of the prediction algorithm was built with a “naive” approach. Roughly speaking, the table prescribes equal proportions for all normal potency scents except for variable potency scents. Employees are expected to develop a sense of composition beyond the simple guidelines in the manual. The first prediction algorithm merely implements the manual guidelines, as an employee on day-one would be expected to do, without any experience-based intelligence. Since the ultimate goal of the prediction algorithm is to incorporate data from previous successful formulas into future ones, the naive algorithm will likely only be used for comparison purposes.
The problem of algorithmic formula composition is fundamentally a parameter optimization task. The structure of the data makes it difficult to come up with a meaningful experimental design, however. Each formula is a piece of categorical data which contains any number of other categorical items (ingredients). The parameter that must be optimized is a numerical quantity associated with each ingredient. The standard technique of dummy-encoding categorical variables might impose an arbitrary and incorrect ordering. For this reason, a standard statistical analysis is not being used.
The second iteration of the prediction algorithm uses a stochastic optimization algorithm. Stochastic algorithms introduce small, random changes and keep those that are beneficial. Stochastic algorithms are uniquely suited for this problem because they can search a very large solution space with no prior domain knowledge. This is important because there is no mathematical procedure for codifying intuition. We can express a formula as a set of parameters, however, and maximize the degree of similarity between predicted and actual formulas. The parameters being optimized are a scaling factor associated with each scent and a compatibility factor between scents.
Justification
It is believed that there have not been any previous attempts at algorithmic perfume formula composition, so there is no status quo by which to compare the performance of the system.
The prediction algorithms work by incrementally minimizing the average percent error between predicted formulas and formulas in the training set.
The system functions to eliminate the need for each employee to learn how to manufacture the custom fragrances.
System Model
Employees will have access to the formula composition interface 1207. They will also be allowed to search for previous formulas in block 1205 by the formula ID or the owner's last name, compose formulas, and collect demographic information from the customer in block 1209. At the end of this process, the information will be uploaded to the database in block 1211.
Administrators will be able to access all of the employee functions described above, as well as those blocks having boldface text in
Functional Requirements
Employees
One benefit of this system for employees is the streamlining of the customer service process. As shown in the system model of
If the customer is reordering in step 1205, the employee may navigate to the desired formula by either searching for the customer's last name or by directly inputting an identification number associated with that formula. When searching by name, the application will display all of the formulas created by that customer and allow the employee to choose one.
When creating a new formula in block 1207, the employee will input the scents selected by the customer to be included and the software subsystem 1200 will automatically generate a composition based on previous formulas. The employee can then adjust the amount of each scent based on customer preference before moving on to the final step.
Once a formula has been created or selected, the employee will enter the customer's demographic information; this includes name, address, birthday, customer notes, and a formula name if the customer so chooses in block 1209. The subsystem 1200 will ask for confirmation before updating the database; at any point before this step, the user can go back to a previous step to make changes.
Administrators
Administrators must have access to the same functions as employees, but also have additional functional abilities which are prompted to the Administrator in block 1213. These allow the Administrator the ability to directly interact with the database in block 1217. They will be able to select from four broad categories of data:
The administrator will be able to add, delete, or update entries in each of these categories.
It is also the administrator's responsibility to run the prediction algorithm in block 1219 which determines the alpha and beta coefficients for the generalized equation. The prediction algorithm is also run after adding new formulas to the training set. The algorithm will automatically terminate once it detects that improvements are no longer significant, or can be manually stopped when desired.
The administrator has the ability to search, view and display the scents most often included in customer formulas in block 1215. Various other search criteria may also be used.
User Interface Specification
Each block in the system model corresponds to a different screen and sub-task in the application:
When creating a new formula in block 1207, the user will first be presented with a tabbed menu of scents.
There will be at least three tabs (top, middle, and base) representing one of the many ways in which scents can be categorized; these particular categories correspond to scent molecular weight and the length of time a scent remains detectable after being applied. Typically, a balanced formula contains at least one scent from each category. Scents chosen by the customer will be selected on the screen by the employee via check boxes and the user will then proceed to the adjustment screen.
The adjustment screen will display the algorithmically generated drop count for each scent selected in the previous step. The user can modify these numbers up or down via arrow buttons or can directly change the value by typing inside the text box that displays the drop count.
Database Modification Interface
In block 1217, the employee will select one of the four information types (employee, customer, formula, or scent) from a drop-down menu. The screen will then display the database contents for that category in the form of a spreadsheet-like grid of text boxes. The employee can add a new entry by clicking the “add” button, or delete entries by selecting a checkbox at the beginning of each row and clicking a “delete” button. When the user is done editing, they will click the “save” button to update the database.
While the present disclosure illustrates various aspects of the present teachings, and while these aspects have been described in some detail, it is not the intention of the applicant to restrict or in any way limit the scope of the claimed systems and methods to such detail. Additional advantages and modifications will readily appear to those skilled in the art. Therefore, the teachings of the present application, in its broader aspects, are not limited to the specific details and illustrative examples shown and described. Accordingly, departures may be made from such details without departing from the spirit or scope of the teachings of the present application. Moreover, the foregoing aspects are illustrative, and no single feature or element is essential to all possible combinations that may be claimed in this or a later application.
This application is a Divisional of U.S. patent application “SYSTEM FOR CREATING CUSTOM FRAGRANCES” Ser. No. 14/936,662, now U.S. Pat. No. 10,242,391, filed Nov. 9, 2015 which claims the benefit of U.S. Provisional Patent Application “SYSTEM FOR CREATING CUSTOM FRAGRANCES” Ser. No. 62/078,944 filed Nov. 12, 2014 and hereby incorporates these by references in its entirety.
| Number | Name | Date | Kind |
|---|---|---|---|
| 3696161 | Kobetz et al. | Oct 1972 | A |
| 5031764 | Meador et al. | Jul 1991 | A |
| 5626155 | Saute | May 1997 | A |
| 6402120 | Swaab | Jun 2002 | B1 |
| 6516245 | Dirksing et al. | Feb 2003 | B1 |
| 6800015 | Derges | Oct 2004 | B1 |
| 10242391 | Fleming | Mar 2019 | B2 |
| 20030014324 | Donovan et al. | Jan 2003 | A1 |
| 20070191257 | Andretta | Aug 2007 | A1 |
| 20080027820 | Brill | Jan 2008 | A1 |
| 20080131858 | Gordon | Jun 2008 | A1 |
| 20090068068 | Harris et al. | Mar 2009 | A1 |
| 20090156501 | Braunn | Jun 2009 | A1 |
| 20100049636 | Sawada et al. | Feb 2010 | A1 |
| 20140032356 | Chuang et al. | Jan 2014 | A1 |
| 20150182578 | Cavallaro | Jul 2015 | A1 |
| Number | Date | Country | |
|---|---|---|---|
| 20190172108 A1 | Jun 2019 | US |
| Number | Date | Country | |
|---|---|---|---|
| 62078944 | Nov 2014 | US |
| Number | Date | Country | |
|---|---|---|---|
| Parent | 14936662 | Nov 2015 | US |
| Child | 16267117 | US |
