PVDF is used also sensory material. And the amplitude Fig. Tanakaof the output is evaluated. The sensor is attached to the tip of finger and a small-sized motor is attached to the base of the finger to excite the sensor to the object. When the object is harder, the amplitude of the sensor output is larger. We tried to measure the various prostates in clinical tests, such as normal prostate, prostatic cancer, hypertrophy, with stones.
The hardness of prostatic cancer is similar to that of bone, and that of prostatic hypertrophy is elastic. Concerning the prostate with stone, the hardness of the part of stone is as same as stone.
Emotional Engineering (Vol. 3)
The prostate con-ditions of subject are as follows. Normal and healthy, B: It is seen that the outputs on subjects A and B are much smaller than the others and the output of B is slightly larger than that of A. The difference corresponds to the difference of the disease A and B. The sensor value on subject C is much larger than that of subject A and B. It means that the state of the prostate is closer to the hypertrophy. It is seen that the sensor out-put on subject D, who is under treatment of prostatic cancer, is closer to that of subject C.
About the condition of subject D, the palpation result of the doctor without sensor could not distinguish whether the stiffness is that of the prostatic cancer or hypertro-phy. The result of the sensor output means the condition is closer to that of prostatic hypertrophy. The sensor output is effective to evaluate the disease conditions. Subjects E and F have prostate stones in places, and the sensor output takes the maximum and much larger than the others. Subject G has one prostate stone and it was difficult for the doctor to search for the part.
Therefore, the measurements were done many times and the fluctuation of the sensor output was large. From the doctors diagnosis, the state of prostate except the stone part is prostate hypertro-phy. Therefore, the average value became the smaller than that of subject E and F. However, it is noticed the maximum value on subject G is large and the result means that it is unhealthy condition.
Further, the conditions of subjects were investigated using the ultrasound tomography. The enlarged prostate conditions of subject C and D were observed. However, the white mark could not be discovered on the prostate of subject G, by ultrasound tomography. These results showed that the output of the present sensor varies with the stiffness of prostate glands and the present sensor output has a good correlation between doctors palpation result.
Further, it is said the sensor is effective in diag-nosing the condition of prostate glands. In medical welfare fields, the highly accurate palpation sensor is expected to be effective to find the part of disease in early stage and to keep the health. Further, to know the principals and mechanism of the tactile sensation leads to the development of the technology, for example, the tactile display system that can transmit someone else touch feelings and tech-nology that can give reality using of characteristics of touch feelings.
In future, the range to use the technology of the touch feelings will extend more and more. Shepherd GM Neurobiology. Oxford University Press, Inc. Proceeding of 11th sensory test conference, Union of Japanese Scientists and Engineers 3. Yoshikawa Y Mechanical behavior of skin and measurement way hardness measure-ment of a living body and artificial judge. Meas Control 3 Fung YC Biomechanics: Springer, New York 6. Maeno T, Kobayashi K, Yamazaki N Relationship between the structure of finger tis-sue and the location of tactile receptors.
Copyright by Elsevier 8. Tanaka M Measurement and valuation of touch sensation: Stud Appl Electromagnet Mech Tanaka M, Numazawa Yu Rating and valuation of human haptic sensation. Int J Appl Electromagnet Mech J Intell Mater Syst Struct Tanaka M Development of tactile sensor for monitoring skin conditions. J Mater Process Technol Skin Res Techonol 9: Sixteenth international conference on adaptive structures and technologies, pp DEStech Publications Inc.
Smart Mater Struct 9: One critical challenge in emotional design is the measurement and prediction of affect. Most current measurement and prediction methods are affected by many biases and artifacts. For example, verbal reports only represent the sheer reflection of consciously experienced feelings. This study aimed to eval-uate affect via physiological measures. Specifically, standardized affective stimuli in both visual and auditory forms were used to elicit different affective states 7 types of affect for the visual stimuli and 6 for the auditory ones.
Each affec-tive stimulus was presented for 6 s and a wide range of physiological signals were measured, including facial electromyography EMG zygomatic and corrugator muscle activity , respiration rate, electroencephalography, and skin conductance response SCR. Subjective ratings were also recorded immediately after stimu-lus presentation. The physiological measures show a relatively high differentiating ability in postulating affect via statistical tests and data mining-based prediction, with highest mean recognition rates of This technological and methodological advancement offers a great potential for the development of emotional design.
Jiao The George W. Emotional design capitalizes on this perspective by conceptualizing affect-engendering products, by adapting to and by responding to human affective states to design products desirable to humans holistically. For example, products are now designed for hedonic pleasure [3], emotional responses and aspirations [4], and user experience [5], and so on, to improve customer satisfaction.
In some instances, optimal performances require an appropriate arousal level [6], such as in aviation safety and repetitive work where a state of high vigilance is desirable [7]. Csikszentmihalyi [8] proposed the concept of flow that it is often necessary to maintain or prevent partic-ular affective states to have optimal performance and enjoyable user experience, such as in tutoring and training, driving, and video gaming.
Although the importance of affect as a design parameter has been well recog-nized, several fundamental questions regarding affect acquisition, measurement, and evaluation remain not well-answered: Research in psychology sheds light on solving these questions. Psychologists often frame affect in two approaches: The former conceptualizes affect discretely. For example, Ekman [9] devises a list of 15 basic emotions, among which 6 are primary emotions sadness, happiness, anger, fear, disgust, surprise.
Research on the latter has found that the most commonly used dimensions are autonomic arousal sleepy-activated and hedonic valence pleasure-displeasure [10], such as Russels circumplex model [11]. As for the first and second questions, current practice in emotional design often uses subjective methods retrospectively in terms of affective adjectives in a discrete manner, such as user interviews, focus groups or surveys [12].
A large amount of affective adjectives are first collected concerning the consumers feelings toward a product or other affective stimuli. Then, the most relevant and appropriate terms are selected by domain experts, ranging in numbers from several dozens to several hundreds. The selected ones are further scrutinized and structured, either manually or statistically and evaluated based on n-point Likert scales e.
These methods are convenient, and amenable to statistical analy-sis. However, verbal account of feelings only captures part of affect, and is usually expressed in abstract, fuzzy, or conceptual terms [14]. Hence, work on affect elic-itation and acquisition is often based on vague assumptions and implicit inference. Moreover, they often suffer from recall and selective reporting biases if affective responses are reported retrospectively [15]. Although the affective information can be collected concurrently, the biggest problem is their interference with the task or activity when eliciting and acquiring affective responses [10].
With regard to the third question, besides the subjective methods mentioned above, another direct method is to appraise what the user is feeling from observing their non-verbal and verbal Eliciting, Measuring and Predictingexpressions and reasoning their situations [16]. However, observation is often costly, time-consuming, and sometimes disruptive for many design tasks [17].
Evidence has shown that physiological signals can differentiate basic emo-tions [18]. In addition, physiological data can be acquired in a continuous manner which is consistent with the way people perceive emotions [20] and user affective states can be evaluated in real time.
PriceMinister - Erreur
Compared with subjective self-report methods, the real-time feature is particularly useful for prod-ucts that seek to respond to a users affect in ways to improve the interaction [21]. For example, Mandryk and Atkins [22] employed psychophysiological techniques to evaluate video games and proved that directly capturing and measuring autonomic nervous system activity could provide accesses to user experience. Therefore, the user-product interaction can be tailored to the individual level to optimize the user pleasure and efficiency [23]. However, it is also suggested one physiological measure alone is not adequate to give a coherent picture of what affective state is occurring within the user [24], since the relationship between psychology and physiology is not all one-to-one and in many cases, many-to-one, one-to many, or many-to-many [25].
Thus, it is necessary to be able to monitor multimodal physiological signals. The purpose of this study is, to what extent, multiple physiological signals can be used to acquire, measure and predict user affect with standardized affec-tive stimuli in both visual and auditory forms in real time.
- There was a problem providing the content you requested!
- Emotional Engineering vol. 2 ||!
- Verfangen: Im Netz des Bösen (German Edition).
- Towards Democratic Transitions: The Case of North Africa;
- Books by Shuichi Fukuda.
- The Bunklers Story.
- What is Kobo Super Points?.
For one thing, evidence has shown that both types of stimuli can be used to evoke affective reactions in physiology [26, 27]. Moreover, visual and auditory information is the most fre-quently used stimuli presented in products and systems that might elicit affect. Also studied is the question whether there is any difference in affective responses when participants are exposed to different forms of affective stimuli. In order to do so, statistical tests and data mining methods were employed.
Initial results from statistical rests showed its potential while affect prediction based on data mining methods delved into a more detailed and specific results for further proving the potential of the proposed method. In order to increase homogeneity within cultural groups, Chinese and Indian students were required to be born and raised in mainland China and India, respectively and to have lived in Singapore for less than 2 years; while the Westerners were exchange 44 F.
Jiaostudents at Nanyang Technological University, Singapore, for less than six months. Each participant was interviewed by the experimenter to screen out mental disorder or drug and alcohol abuse. Due to technical difficulties, only data collected from 14 par-ticipants in each cultural group i. Informed consent was obtained from each participant. Based on the participants reports, the visual and auditory stimuli were categorized into seven and six groups, corre-sponding to different affective states respectively. The physiological sensing system included an 8-channel Biofeedback and Neurofeedback System v5.
The latter were used to collect facial EMG signals zygo-matic and corrugator muscle activity. The E-Prime software was used to ensure millisecond precision data collection. Besides, auditory stimuli were presented to the participants through a pair of Altec Lansing AHP headphones. The sensor placement on the participant is shown in Fig. It has been proved to be linearly related to affective arousal and independent from affective valence [26].
SCR in s is measured by fastening the electrode straps around the second and the fourth fingers of the participants left hand. Four temporal features of SCR were computed in Fig. Facial EMG measures muscle activity by detecting and amplifying the tiny electrical impulses that are generated by muscle fibres when they contract [31]. Facial EMG measures from the zygomaticus major ZM and corrugators supercilii CS are widely used for recognition of affective states [32].
EMG signals in V are bandpass filtered from to 1, Hz using an elliptic, short infinite impulse response IIR filter, amplified 5, and rectified [26]. The baselines for ZM and CS are defined as the mean activity in the one second before stimulus onset. Then, the change scores for both ZM and CS were computed by subtracting their respective baselines from the mean responses sampled at each half-second time interval after stimulus onset for the following six seconds i.
The standard deviations of the raw change scores3. The means of the absolute values of the first differences of the raw change scoresLet Yi refer to the normalized signal, i. The means of the absolute values of the first differences of the normalized raw change scores 4. The means of the absolute values of the second differences of the raw change scores6.
The means of the absolute values of the second differences of the normalized raw change scoresRespiration measures the rate or volume of air exchange in human lungs by its rate or amplitude. It is observed that high arousal generally increases respiration rate while low arousal decreases respiration [22, 34]. In this study, the respiration sensor was sensitive to stretch and was strapped around each participants abdo-men. The respiration rate and relative respiration amplitude data were collected. Empirically, the deviation respiration rate from the baseline was calculated by subtracting the mean rate of 3 s since stimulus onset from the mean rate in 0.
Then the mean, the standard deviation, and the average acceleration or deceleration of the respiration measure were computed from the sampled data as features for model construction. Note the average acceleration or deceleration of the respiration measure is defined as follows: EEG is the recording of electrical activity along the scalp produced by the fir-ing of neurons within the brain and reflects correlated synaptic activity caused by post-synaptic potentials of cortical neurons [35]. It is the summation of oscil-lations with a frequency range from 1 to 80 Hz, with amplitudes of 0.
In this research, the alpha Hz and beta Hz frequencies are extracted to reflect different affect related activities [37]. In this research, the only sensing electrode is placed at the Fpz which is defined by the international system [38]. The two reference electrodes are located at the left and right ear lobes.
In order to minimize artefacts introduced in the EEG sig-nals, participants are instructed not to blink during the six-second stimulus presen-tation as they could, and electrooculogram artefacts are corrected by an adaptive filter based on the least mean squares regression [39] using EEGLab software based on Matlab Rb.
Then the following energy features are calculated: The power of alpha E ,2. The power of beta E ,3. Then, the corresponding energy features are obtained within the specified i. These features tell at which frequency ranges variations are strong and at which frequency ranges variations are week. Before the experiment, a practice trail with 6 pictures and 4 sound clips these stimuli were different from those in the real experiment was con-ducted by the participant.
This enabled the participant to be familiar with the experiment protocols. After all the sensors were installed on the participant, a 2 min resting baseline was followed before the first stimulus was presented on a inch HP desktop around 1 meter away. The experiment took place in a project room with dim lighting at 25 C. The order of stimuli presentation was randomly but not repeatedly generated for each participant to minimize order and habituation effects. After receiving one stimulus, the participant was asked to rate the stimulus on valence and arousal using the computer-version SAM.
Afterwards, each participant was asked to choose one affective adjective that best describe the affective response to the stimuli from the given terms see Sect. To avoid conflict between two consecutive affective states, there was a 10 s rest interval without any stimulus before the presentation of the next stimulus.
The corresponding pictures in the IAPS were identified as 4,, 4,, 4,, 4, excited , 1,, 1,, 2,, 2, Each received mean votes of In order to better differentiate among positive, neutral, and negative affect as well as different affective states, F-values of the 12 sampling data were first cal-culated except subjective rating, SCR, and EEG as there was only one sample for them in the corresponding timeline after stimulus onset and the physiological sampling data associated with maximal F-values and p-values smaller than 0.
First of all, affect was tested based on valence, i. Valence ratings had the most significant main effect, as expected; Post-hoc comparisons between any two valence categories were significantly different, p 0. It is also not uncommon for the final crisp value to lie under multiple curves. In this case, the objective being studied may be considered to belong to intended for more than one target groups. A schematic diagram to summarize the structure of Fuzzy Logic System is illustrated in Fig. As indicated by Fig. Each fuzzified MFP is regarded as the antecedent of the system and multiple ante-cedents are fired through the applications of logic operators, Fuzzy rules and implica-tion method.
Once the output of each MFP and Fuzzy rule is generated, aggregation method then combined all the outputs based on the principle of maximum and turned these outputs into single final MFP. The final MFP is then defuzzified via application of centroid method. This is how qualitative data can be translated into final quantita-tive crisp value, hence indicated the belongings of the object being studied. This section will discuss how the FuzEmotion system was constructed and how these cellular phone features are accessed using such system to confirm the gender inclinations of cellular phones.
Previously, four classes of features being dimension, Ratio A, Ratio B and Mass were pre-processed based on ten sub-attributes Table 6. These sub-attrib-utes were normalized expect mass before turning into four different attributes. The overall quantity of these classes and sub-attributes may leads to complication when assessed using FuzEmotion, hence it is desirable to consolidate these classes and sub-attributes and turned them into super-classes. In this case, both Size and Ratio super-classes were defined to consolidate these classes.
This was achieved through the application of weighting factors Table 6. The topic on Neutral Network is beyond the scope of this chapter, which will not be discussed here. Consolidation of these ten attributes were achieved and defined by equations Eqs. Weighing factors indicate the influence importance of each sub-attribute on the two super classes of Size and Ratio. The new governing equa-tions with weighting factors applied are shown as Eqs.
Similarly, the newly added sub-attributes Table 6. The external cover design of cellu-lar phones were turned into five different classes of top, bottom, keypads, func-tion key and lastly body shapes. Each class of attribute is associated with five 6. As shown in the table, different weighting factors were assigned for each sub-attribute and may not be the same for different gender groups. This was due to the facts that even the same sub-attributes may have different influence on the classes for different gender groups. For example, sub-attribute Top1 is more important for the female Top class w.
These weighting factors were calculated based on degree of appearance or fluctuations of sub-attributes. For example, as a particular sub-attribute is adopted much more than rest of the sub-attributes for designing new cellular phone i. Due to complexity of these classes and sub-attributes, consolidation was con-ducted once again to unify these classes into single super-class.
The new-super class was given name as Form, which defines the external cover design of cellu-lar phones. The new super-class of form and its sub-classes are listed in Table 6. As shown in Table 6. These weighting factors were also calculated based on the degree of appearance or fluctuations of each sub-class. This was looking at the overall influence of individual class Top, Bottom etc. The overall structure of this process can be summarized by Fig.
Note that there are some zero weighting factors such as the weighing factors for Keyad5 and Funckey6, meaning that these sub-attributes were not adopted at all for a particular class and should not have influence on the form design regardless of any gender groups. The new super-class of Form is now defined by Eqs.
All the three sets of Fuzzy inputs were then fuzzified via membership functions and represented graphically with MFPs. The formation of individual MFP is discussed in the following sections. Each set of antecedents are assigned with three sub-categories, where each of the sub-categories of data was fuzzified by one membership func-tion. For Size feature, the sub-categories were assigned as Small, Medium and Large. Similarly for Ratio feature, Low, Medium and High subcategories were assigned.
Lastly for Form feature, the sub-categories were assigned as Form 1, Form 2 and Form 3. Each of these sub-categories was associated with a membership function, which used to transform these sub-categories into MFP. The membership functions for each sub-category are listed as Eqs. The number of Fuzzy Rules required is related to the total number of antecedents and the sub-attributes used in the system. In this case, three antecedents were established being Size, Ratio and Form. Each of the antecedents was assigned with three sub-categories such as Form1, Form2 and Form3. These nine sub-categories have generated total of 27 Fuzzy rules, which are listed as follows: It is a process involving application of Logic Operator and determining the minimum of result from each antecedent fired.
It is difficult to present all 27 sets MFPs here due to space limita-tion, but a simple illustration of single implication method is presented in Fig. Note that the resulting degree of membership is 0. Since AND logic operator is applied alongside of implication method, the minimum of the three resulting degree of membership is taken as the output for this particular set of MFPs Fuzzy Rule This can be repre-sented by Eq. The final shape of gender MFP was initially approximated. Since all the antecedents MFPs Ratio, Size and Form were established based three membership functions and taken the shape of trapezoidal, it is reasonable to assume the final membership function plot must have adopted similar kind of shape.
The OR logic operator was used to determine the maximum of the outputs. Since all the Fuzzy rules are fired in parallel, it likely to have similar output from each set of MFPs but with different degree of membership. Defuzzification is the last step in the construction of FuzEmotion, where the final MFP was defuzzified into single crisp value. This value can be used by cellular phone designer to assess the gender inclination of mobile phones. As mentioned earlier, several methods could be adopted to defuzzify the final MFPs such as centre of mass or centre of maxima.
For simplicity, centre of mass is applied to deal the problem presented in this project. Centre of mass method worked upon determining the centroid of the area under the curves. All cellular phone have been defuzzified into final crisp values. The overall accuracy of previous FuzEmotion system and the current FuzEmotion system are presented in Table 6. It could be expected that upon tuning, the accuracy of FuzEmotion system could be improved. However, it is also reasonable to state that with The solution of FuzEmotion system is a final membership function plot that is integrated with 27 different sets of MFPs and Fuzzy Rules.
This change may be due to two reasons. Firstly, there are more gender neutral phones in the selected database i. Secondly, it was possible that the most current cellular phone design may be intended for gender neutral groups, hence allowing these phones to gain benefits and attentions from all the three gender groups. Note that these values on the x axis are there only for the ease of calculating the centroid of the result.
There is no doubt that functionality of the products has always being one of the important considerations when the end users decide what product to purchase. However, as technologies has pushed the design of cellular phones to the limit with relatively minimised development time. The philosophy of manufacturer-orientated design is no longer effective in the current market of cellular phones. End users are not only satisfied by the functionality of cellular phones, but unique factors such as the appearance of the phone, the style of the phone and lastly good feeling or Kansei about the phone.
Cellular phone manufacturers and companies need to look at this issue from a totally different perspective. Questions such as What are the actual factors that gain the attentions of end users? KE is a concept that aimed to solve this type of abstractive emotionally associated issues by assessing customers true perceptions and integrate these assessed data into product design process, hence achieving market success. The needs from customers are satisfied both emotionally and functionally by KE integrated products. This project focused on the study of cellular phone features and how these features are emotionally connected to end users.
This was achieved through the development of FuzEmotion system, which is Backward KE tool that could be used to assess and confirm the gender inclination of mobiles with an overall accuracy of In the word, FuzEmotion system will be able to tell the users which gender group female, gender neutral or male a particular cellular phone belongs to.
The whole FuzEmotion system consisted of 27 sets of antecedent MFPs, 27 Fuzzy rules and 1 final mem-bership function plot. The system was constructed based on specifically selected phones from a database containing total of phone samples. These samples were extracted from officially released source published on the World Wide Web.
These cellular phones were designed and manufactured by the top five cellular phone com-panies being Nokia, Samsung, Sony Ericsson, Motorola and lastly LG, which well represented the current cellular phone market. Throughout the development of FuzEmotion, several key factors were discov-ered. For example, the weighting factors used to generate MFPs had significant Fig. Note that the current system tended to shift towards male side. This phenomenon may due to two major reasons.
Firstly, the system might has witnessed the trend of current cellular phone design as recently released phones are intended for male or gender neutral customers. Secondly, it was also likely that such change was affected by the quantity of gender neutral phones in the database. Regardless of the difference, FuzEmotion has prove its capability to assess and confirm gender inclination of cellular phone and may acted as a design supporting tool for cellular phone designers. This project has achieved, in part, the two objectives proposed in the earlier sec-tions. The overall accuracy of previous FuzEmotion system has improved from 76 to Although FuzEmotion was not directly related to KE, the concepts behind both systems are relatively similar, which both aimed to study the feeling of customers.
It is reason-able to state that the application KE in product design process should be a feasi-ble approach based on the study of FuzEmotion conducted. Although KE has been existed for almost 40 years, there is no doubt that significant amount of studies are still required until the function of KE system is fully utilized. Nagamachi M Kansei engineering. Nagamachi M Kansei engineering: Int J Ind Ergon Nagamachi M A study of emotion-technology. Japan J Ergon 10 4: Nagamachi M Kansei engineering as a powerful consumer-oriented technology for product development.
Appl Ergon 33 3: Canny J A computation approach to edge detection. Trans Pattern Anal Mach Intell 8 6: Zadeh LA Fuzzy sets. Comput Electron Agric The chapter described a system called ProEmotion for the purpose of assessing the Kansei aspects of a product by con-sidering design attributes of a product. Neural Network is used to process Kansei words. The system has been successfully implemented to ascertain gender inclina-tion of a mobile phone. Principal parameters of a mobile are considered, that is, length, width, thickness and mass.
This is based on a set of 92 mobile phone samples from the five major mobile phone manufacturers. As well as consumers requirement for the needed functionality, consumers feel-ing toward a product has also become an important factor in their choice of the product. Kansei engineering, also known as Kansei ergonomics or emotional engineering, aims at effectively analyzing and incorporating customers feeling and inclination into product function and product design.
Kansei engineering was founded in Japan in the late s [13]. An example of emotional connectedness would be Chapter 7ProEmotion: This feeling is called Kansei in Japanese. In other words, Kansei means the customers psychological feeling as well as embracing physiological issues [8].
Kansei words are the expressions of the attributes of Kansei aspects. A Kansei engineering system is often made up of a computer-assisted system of Kansei engineering, an expert system and databases []. Kansei words are input into the system and recognized in reference to the Kansei word in the data-bases. The words are matched to the image database and calculated by an infer-ence engine to find the best-fit design, which may be shown for example on the display of a computer. It is also very useful to assist a designer to better understand the new product being developed [8]. Such a system supports a flow from Kansei to the design domain and hence is called forward Kansei engineering forward KE.
These Kansei findings can then be used for product improvement or new product development. This is called backward Kansei engineering backward KE. Today, applications of Kansei engineering have been found in a number of industries such as automotive, con-struction, machine tools, electric home, costume, cosmetics [8, ]. Many modern electronic products such as mobile phones have reached their functionality peak for most of the mobile phone users.
Consumers are slowly shifting their focus from functionality to fashion. Young people in particular want their phones to be unique, representative of personality and symbol of fash-ion. Consequently, modern day consumers are a lot harder to be satisfied when it comes to personalized electronic goods.
One major effect on mobile phone manu-facturers is the demand for an increased pace of new product introduction because the mobile phones shelf-life has been greatly shortened. A wide range of new mobile phones are being designed and pushed into the market to meet a population of diverse customers. However, there are few quantitative methods the companies can use to ascertain whether their new models appeal to the targeted consumer. Reviews have indicated that there are mobile phones in the market that were intended toward a specific group of consumer but in fact failed to meet the consumers expectations [29, 30].
A Tool to Tell Mobile Phones GenderIt is evident that modern mobile phone manufacturers are in a matured market in terms of availability, technical functionality and cost, but there is a need to pay greater attention to the Kansei emotional or affective aspects of their products. In other words, the companies need to ask questions such as, a what makes a phone more appealing to a male as against a female?
This is effectively a backward Kansei engineering exercise Fig. This chapter presents a Neural Network method for appraisal of mobile phones gender inclination in terms of male, female and general public i. The product attributes considered are only the principal ones, namely length, width, thickness and weight of a phone. Choice of these attributes is discussed in the following section. A total number of modern mobile phones from some of the major brands i.
The information sourced is that of the officially released product specifications http: Based on the discussions earlier, only principal attributes, that is, dimensions and mass, are considered in this research. The relative ratios of keypad and screen size to phone length are two attributes that are commonly believed to have a strong bearing on Kansei aspects of a phone.
This is true across the board of different brands and customer groups. Both functionality and price information are not considered as they have little bearing on a phones Kansei. Neural Network is chosen as a solu-tion toward modeling the problem. The mobile phone database provides ample input data, which is supposedly to be mapped to a specific output, that is, gender classification. The key is then to figure out the explicit numerical transformation functions in-between, which in many cases may or may not exist.
Due to the abil-ity to learn from data like a human brain, one of the advantages of Neural Network is its ability to work with raw data alone without requiring deeper and thorough understanding of the process in the process of obtaining the required result. Other advantages include being robust toward noisy data and its broad application in many abstract classification problems. In essence, Neural Networks are trained to solve problems based on generat-ing functions to correlate between input data and desired output.
Given a particular input x, Neural Network generates func-tions to achieve the output y. Function f x could be defined as a composition of other functions gi x , which could once again be embedded into other functions hi x. The relationships between these functions could be visualized as a network structure Fig. ProEmotion is a software tool based on the Matlab Neural Network toolbox.
This software considers the database of mobile phones and their targeted gender market to give opinions on mobile phones. ProEmotion can also produce and plot the accuracy of functions, give importance of each parameter, display similar phones to a given phone and save results to a file.
Samenvatting
Therefore, feature data were listed on a spreadsheet-like format with no labeling for rows and columns. Target audience for each phone is imported into Neural Network as a separate CSV file with corresponding arbitrary values for Male and Female audience. First, the appropriate functions are defined. Matlab has three unique Neural Network toolboxes which can be used in different situations. First of all, the func-tion-fitting tool is used to predict a certain functional trend. Secondly, the Pattern Recognition tool helps classify specific data.
Finally, the Clustering Data tool is used to group data by similarity. For the purpose of this research, the Pattern Recognition and Clustering Data tool was utilized to achieve the final result. After the right functions are selected, Neural Network requires an input and target output for the system to train upon. This input for this research was a database of existing cellular phones in the mar-ket. Meanwhile, the target output was to identify which samples within the data-base were their respective genders according to market research.
After the appropriate data are imported, the training parameters need to be setup. The user may assign a set of samples to be trained, validated and tested. Training sam-ples are used as means to generate functions for Neural Networks learning process. Validation samples are used to also generate functions and check against the Training functions. Testing samples are used to test whether the function network is accurate by comparing its results with the target output. The user also needs to assign the number of neurons for training, which acts similarly to brain cells.
However, it is worth noting that more brain cells do not necessarily mean better results. For optimal results, the number of neurons should be adjusted according to the number of sample inputs.
Once the training is completed, the user is able to read the output data and graphs, and determine whether the system is satisfactory or not. If the system is unsatisfactory, there are several ways to adjust the system. The user may choose to retrain the network, hop-ing for a different accuracy turnout, though this may not be effective if the accuracy was very poor in the first place. The user may choose to adjust the training param-eters i. The distribution of samples for Training, Validation and Testing can affect the results.
Bedankt voor uw beoordeling Uw reactie is inmiddels op de website geplaatst en zal door onze redactie beoordeeld worden. Lezersrecensies 1 Vond u deze recensie nuttig? Contactformulier Vult u alstublieft onderstaande gegevens zo volledig mogelijk in, dan kunnen wij u optimaal van dienst zijn.
Wij nemen binnen een werkdag contact met u op. U kunt ons ook bellen op Hieronder kunt u uw verzoek toelichten of een vraag stellen. Bedankt voor uw aanvraag Wij nemen zo spoedig mogelijk contact met u op. Download brochure Vult u alstublieft onderstaande gegevens in om onze brochure in PDF-formaat te downloaden. Het downloaden start direct na het verzenden van deze gegevens.
Structure and Form in Design. Games and Learning Alliance. Science, Technology, and Applications. Advances in Usability and User Experience. Advances in Neural Networks. Design for Teaching and Learning in a Networked World. Collaboration in Creative Design. Information and Knowledge Design. The Neuroscientific Basis of Successful Design. Human Aspects in Ambient Intelligence.
- The Mysterious Mr. Miller.
- The Muleshoe Connection (Not To Be Confused With the French Connection).
- Emotional Engineering (Vol. 3) (eBook, PDF).
- Emotional Engineering - Google Книги.
- Join Kobo & start eReading today?
- ?
Style and Creativity in Design. Advances in Human Factors and Sustainable Infrastructure. Computer Science in Sport. Computational Intelligence and Decision Making. Comparative Design of Structures. Emotions and Personality in Personalized Services. The History of Visual Magic in Computers. Human and Robot Hands. Smart Education and e-Learning Expanding the Frontiers of Visual Analytics and Visualization.
Human Walking in Virtual Environments. Learn Autodesk Inventor Basics. Human-Harmonized Information Technology, Volume 1. New Directions in Concurrent Engineering. Advances in Affective and Pleasurable Design. How to write a great review.