Effect of Feedback Medium for Real-time Mastication Awareness

Increase using Wearable Sensors

Guillaume Lopez, Hideto Mitsui, Joe Ohara and Anna Yokokubo

Graduate School of Science and Engineering, Aoyama Gakuin University, Tokyo, Japan

Keywords:

Mastication, Bone-conduction Microphone, Eating Quantiﬁcation, Persuasive System, Real-time Feedback.

Abstract:

Increasing the number of mastications can help suppress obesity, but it is difﬁcult to keep constant awareness

of it in everyday life. Besides, the conventional mastication number measurement apparatus is large and non-

portable such it is difﬁcult to use it in daily life. This research proposes a system that supports chewing and

utterance consciousness improvement in real-time. It is composed of a cheap and small bone conduction

microphone to collect sound intra-body sound signal, and a smartphone that processes sound and provides

feedback in real-time so that it can be used conveniently in daily life. First, the accuracy of mastication

counting and utterance length estimation has been evaluated, conﬁrming to be sufﬁcient to provide real-time

feedback. Second, the evaluation of the effect on chewing and utterance consciousness of different ways and

medium of real-time feedback during a meal was carried out. It was possible to clarify the impact of real-

time feedback, as well as to determine the factors that affect more efﬁciently the improvement of mastication

number and utterance.

1 INTRODUCTION

Obesity may cause lifestyle diseases such as diabe-

tes and heart disease. The Japanese Ministry of He-

alth, Labor and Welfare has taken measures for this

prevention, but the number of obese patients has not

decreased compared to 10 years ago (MHLW, 2016).

As measures against obesity, it is known to be useful

to improve eating habits and exercise moderately, but

it is also possible to signiﬁcantly prevent it by incre-

asing the number of mastications (Ando et al., 2008;

Nicklas et al., 2001). As a concrete example, when

attempting to improve mastication activity for young

Chinese men with obesity, it was possible to reduce

the intake of energy in all the subjects consistently (Li

et al., 2011). The same study also demonstrated that

there is a useful possibility of measures against obe-

sity by the activity of increasing the number of che-

wing.

Improvement in the mastication amount is also

crucial since healthcare experts always check the

number of chewing as well as meal duration and food

type as an indispensable factor in assessing dietary

habits. In addition to the above, to prevent obesity,

the nervous system and chewing activities are clo-

sely related. This relation is because chewing repeti-

tion stimulates the satiety center and sympathetic ner-

vous system, which can reduce obesity by secreting

hormones that suppress appetite (Kao, 2007). No-

tably, several past works have reported that people

with fast-eating have higher tendency to be obese,

which is partly because lowering secretion of hor-

mones by eating fast causes an increase in dietary

amount (Denney-Wilson and Campbell, 2008; Gaul

et al., 1975). In addition to this, it is considered de-

sirable to encourage utterances during meals. Indeed,

Kishida et al. have reported that making conversation

during meals is related to good health (Kishida and

Kamimura, 1993).

This research proposes a system that supports che-

wing and utterance consciousness improvement in

real-time. It is composed of a cheap and small bone

conduction microphone to collect sound intra-body

sound signal, and a smartphone that processes sound

and provides feedback in real-time so that it can be

used conveniently in daily life. Though chewing

and swallowing processes depend on many factors

both human and food property dependants (see (Lo-

gemann, 2014)), this paper focuses the attention on

the point that in identical ”food conditions” (content,

amount, etc.) adequate real-time feedback of mas-

tication amount and utterance duration has a positive

effect on the quantitative and qualitative improvement

of both behaviors.

442

Lopez, G., Mitsui, H., Ohara, J. and Yokokubo, A.

Effect of Feedback Medium for Real-time Mastication Awareness Increase using Wearable Sensors.

DOI: 10.5220/0007569804420449

In Proceedings of the 12th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2019), pages 442-449

ISBN: 978-989-758-353-7

2 STATE-OF-THE-ART

Many studies have been focusing on chewing as an

improvement of dietary habits, mainly proposing va-

rious methods and devices to quantify mastication

activity with little burden. As an effort to improve

such nutritional habits, eating habit improvement sys-

tems using wearable devices have been proposed, but

there is still room for improvement in judging mas-

tication amount and utterance duration when used in

everyday life(Amft et al., 2005; Shuzo et al., 2010;

Zhang et al., 2011).

As previous works on mastication counting have

shown, current devices that measure myoelectric po-

tential from the masseter muscle can count bite, but

wearing the apparatus in daily life is a signiﬁcant bur-

den for the user (Kohyama et al., 2003). Another

technique using infrared sensor can detect small chan-

ges in temporal muscle tension, but one can consider

that this method is not applicable in the sense that it

bothers users during meals due to the sensing medium

and the appearance (Obata et al., 2002). Similarly,

Tanigawa et al. explored the use of the Doppler effect

in their system to sense the Doppler signal of masti-

cation produced from vertical jaw movements (Tani-

gawa et al., 2008). However, there also some calibra-

tion is required.

Recently, analysis of internal body sounds spectra

has attracted attention as a way to differentiate bet-

ween biting and speaking activities, and to classify

several types of food with less burden (Mizuno et al.,

2007). Uno et al. Proposed a system to detect the

chewing frequency and bite ﬁdelity using bone con-

duction microphones(Uno et al., 2010). Paying atten-

tion to the amplitude during chewing, it is a system

that judges chewing when amplitude magnitude ex-

ceeds a certain level, and the judgment accuracy was

about 89%. However, activity discrimination method

is limited to speciﬁc ailments. Similarly, Nishimura

et al. (Nishimura and Kuroda, 2008) and Faudot et

al. (Faudot et al., 2010) propose to measure the che-

wing frequency using a wireless and wearable in-ear

microphone. However, to estimate the number of che-

wing operations, still, some parameters need to be ad-

justed by the user each time, which is a severe con-

straint in practical use.

Inoue et al. have proposed a method of overwri-

ting the visual appearance of food and the acoustic

signature of mastication sound during eating using an

head-mounted display (HMD) and a bone conduction

speaker(Inoue et al., 2016). Using the HMD, they

tried to increase the number of chewing by superim-

posing an indication of texture that makes food feel

stiff and regenerating chewing sound which makes it

feel solid in bone conduction speaker. As a result,

they reported that audiovisual information overwri-

ting issued an increase in the number of chewing.

Similarly, Kumagai et al. developed a game as an

opportunity making medium to increase the number

of chewing (Kumagai et al., 2016). Since a personal

computer (PC), an HMD, a charge ampliﬁer, etc were

components commonly used in the above two related

studies, the apparatus was large and difﬁcult to handle

in daily life. In most current systems and approaches,

measurement of conversation time depends on envi-

ronmental sounds, so there is room for improvement

in determination accuracy. Also, even if some works

are doing real-time measurements, feedback display

is not done in real time.

3 PROPOSED SYSTEM

3.1 System Outline

Based on the above review of related research, it has

been decided to use a bone-conduction microphone to

enable the collection of both chewing and utterance

activities information. Some of the hands-free head-

sets available on the market are integrating such spe-

ciﬁc microphone, making it easily accessible to ever-

yone. Moreover, Fontana et al. have shown earlier

that even a strain sensor to detect chewing events and

a throat microphone to detect swallowing sounds pre-

sent enough comfort levels, such the presence of the

sensors does not affect the meal (Fontana and Sa-

zonov, 2013). A smartphone is used to deal with

the real-time processing of the collected sound sig-

nal. The algorithm that counts the mastication num-

ber and estimates the duration of utterance, though

designed based on previous works, has been tuned fo-

cusing on light computation to enable real-time pro-

cessing. Though the accuracy of the proposed system

is not the main topic of this paper, an initial evaluation

has been carried out to conﬁrm it is accurate enough

to give quantiﬁed information about chewing and ut-

terance that is not too far from that perceived by the

user.

The proposed support system has been separated

mainly into two sub-systems that are, a mastication

number improvement support interface, and an utte-

rance awareness improvement device. Figure 1 shows

the usage image of the mastication number impro-

vement support sub-system. In addition to the bone-

conduction microphone (Motorola ﬁniti Bluetooth he-

adset) attached to one ear and the smartphone (Mo-

torola Moto G) to process sound signal, the chewing

frequency improvement support sub-system uses the

Effect of Feedback Medium for Real-time Mastication Awareness Increase using Wearable Sensors

443

same smartphone or a smartwatch (Motorola Moto

360) to provide real-time feedback. Figure 2 shows

the usage image of the support sub-system for ut-

terance awareness improvement. In addition to the

smartphone, it uses a microphone to collect sound sig-

nal and provides real-time feedback about utterance

duration using whether the smartphone display, a vi-

bration device or a Light Emitting Diode (LED). Ar-

duino compatible micro-controllers were used to con-

trol the input of the microphone, and activate whether

the vibration element or the LED simultaneously.

3.2 Method for Real-time Measurement

of Mastication Count and Utterance

Duration

Figure 3 shows the overall ﬂow of the measurement

algorithm for mastication count and speech time. The

main algorithm, though designed based on previous

works, has been tuned focusing on light computation

to enable real-time processing. Short-term energy

was calculated from the raw sound signal and the re-

sulting data used for differentiation of mastication and

utterance signals. Calculation formula of Short-Term

Energy is described in Equ.1, where ”s” is the pro-

cessed raw signal, ”n” the step size in sample num-

bers, and ”w(n)” the windowing function. Short-Term

Figure 1: Usage image of mastication number improvement

support system.

Figure 2: Usage image of utterance awareness improvement

support device.

Energy makes it easier to grasp the characteristics dif-

ferences by explicitly checking the magnitude of the

resulting wave signal. Therefore, by simplifying the

waveform and making the waveform of the sound data

acquired by this system easier to interpret quantitati-

vely, it is possible to capture the features of mastica-

tion and utterance. From the data obtained by Short-

Term Energy, we measured the chewing count and ut-

terance time using both a magnitude threshold and a

duration threshold set by calibration.

E(n) =

liminf

∑

m=−lim inf



s(m) ∗ w(n − m)



(1)

Measurement of chewing count and utterance du-

ration is performed as follows. The outline of the pro-

cess consists in repeating the following succession of

operations in real-time during the meal: sound signal

collection, short-term energy calculation, mastication

and utterance differentiation, result feedback. The

measurement starts after the microphone is set and

Figure 3: Overall ﬂow of chewing and utterance determina-

tion algorithm.

HEALTHINF 2019 - 12th International Conference on Health Informatics

444

connected by pressing the ”start” button of the dedi-

cated smartphone application, and it ends by pressing

the ”end” button of the application. Figure 3 shows

the overall ﬂow of the algorithm for mastication and

utterance activities differentiation.

As shown in Fig. 3, when the calculated short-

term energy value exceeds the threshold value, discri-

mination and calculation of mastication and utterance

are performed. In the case where the calculated values

exceed the set threshold value for more than a ﬁxed

timespan, the algorithm judges the current sound seg-

ment an utterance. The segment ends once the cal-

culated value falls below the threshold, and the algo-

rithm provides the utterance duration to the interface

for real-time feedback before looping again. In the

case where the computed values exceed the set thres-

hold but only within a ﬁxed timespan, the algorithm

judges the current sound segment as a chew and in-

creases the total number of mastications. Then the

algorithm provides this total number of bites to the

interface for real-time feedback before looping again.

Fig. 4 shows examples of the signal after short-term

energy calculation and the timing to separate mastica-

tion and utterance sound segments.

Figure 4: Determination of mastication and utterance and

separation.

4 EVALUATION EXPERIMENT

4.1 Experimental Outline

Evaluation experiments were conducted to not only

verify the accuracy of the above-described algorithm,

but also to clarify the inﬂuence of the different real-

time feedback interfaces and medium on mastications

number and utterance awareness. Table 1 shows the

correspondence of each feedback tested. Patterns A,

B, and C were used in the sub-system that promo-

tes mastication, while patterns D, E, and F were used

in the sub-system that promotes utterance. Figure 5

shows screenshots of the smartphone displays for pat-

tern A and B.

Table 1: Description of the six feedback contents.

feedback content device

A picture animation and gauge change smartphone

according chewing count smartphone

B chewing count smartphone

C pattern A smartwatch

D speech time smartphone

E vibration according speech time vibration unit

F LED lighting according speech time LED unit

Figure 5: Screenshots of smartphone screen’s feedback pat-

terns A (left) and B (right).

An experimental study about mastications count

and utterance duration feedback has been carried out

on 18 males subjects from 22 to 23-year-old. During

each experiment, the subjects ate provided food with

the ear-worn device attached and answered the ques-

tions of the experimenter during meals for utterance

production. Two cases were presented, namely, when

presenting feedback using the proposed system du-

ring meals and not presenting feedback. In both cases,

the ear-worn device was attached. The meal content,

two rice balls and shredded cabbage (60g x 2) were

divided and distributed respectively to nine subjects

each. In the experiment with real-time feedback, sub-

jects have been divided into three groups, such among

feedback patterns A, B, and C, each has been provided

to six subjects respectively. Simultaneously, experi-

ments on utterance time have been performed. Due to

some equipment trouble, only 14 subjects data were

collected. Seven subjects were provided patterns D

and E, and seven others patterns E and F respecti-

vely. However, feedback pattern D had no experi-

mental equipment trouble such the whole 18 subjects

data were available to evaluate the accuracy of utte-

rance duration estimation. Exact values of mastica-

tion count and utterance time were measured using

video recorded by a video camera.

Effect of Feedback Medium for Real-time Mastication Awareness Increase using Wearable Sensors

445

In addition to the above, subjective evaluation

using a questionnaire survey was conducted to eva-

luate whether the user was conscious of the utterance

by feedback on the utterance time. Each question was

answered using the ﬁve-point scale. Furthermore, to

verify whether the user was aware of speaking, Fis-

her’s exact test of the independence between pattern

E and pattern F was performed on the item ”Were you

aware of an utterance during meals?”. Similar sub-

jective evaluation was also carried out in mastication

improvement sub-system experiments.

4.2 Experimental Results

Table 2 shows the results for the real-time detection

accuracy of mastication and utterance sound segments

using the proposed system. Bites could be counted

with sufﬁciently high efﬁciency with an average accu-

racy of 91%. Concerning utterance duration, the over-

all average accuracy was about 96%, which was also

considered high enough to provide feedback in accor-

dance with user’s perception.

Table 3 shows the result of mastication amount in-

crease result when providing feedback compared to

no feedback. An increase could be conﬁrmed for all

18 subjects, whatever the feedback and food types

are. Relatively to each subject, mastication amount

was signiﬁcantly increased (p <0.01, n=18) by about

16% with real-time feedback compared with the case

without feedback. Since the data set has two pa-

rameters, the food type, and the feedback pattern,

a two-way statistical analysis of variance (ANOVA)

has been performed. Table 4 shows the result of the

ANOVA. The p-value of the feedback pattern, 0.02, is

small enough to indicate strong evidence of its effect

on average mastication amount increase. On the other

hand, the p-value for the food type, 0.69, is big, indi-

cating it does not affect mastication amount increase.

Also, since the p-value for the interaction term, 0.89,

is big, indicating the effect of feedback pattern does

not depend on food type.

Then, a multiple comparison test (alpha=0.05,

Bonferroni method) of the mean mastication amount

increase depending on each feedback pattern has been

performed to see if the were any signiﬁcant difference

in the effect of each feedback pattern (5). As a result,

the mean mastication amount increase in case of feed-

back pattern A is signiﬁcantly higher than in case of

feedback B p-values 0.028. The p-value of the mean

difference between feedback patterns A and C, 0.066,

though over 0.05, is small enough to indicate the more

signiﬁcant effect of feedback pattern A. Concerning

the difference between feedback patterns B and C, the

large p-value suggests there is no difference between

them.

Table 2: Chewing count and speech duration detection

accuracy.

chew count speech duration

detection accuracy (%) 91 96

standard deviation (%) 4 3

Table 3: Mastication amount increase result depending on

feedback pattern.

food type subject feedback mastication

pattern increase amount

1 A 46

2 A 31

3 A 32

4 B 29

2 rice balls 5 B 6

6 B 22

7 C 15

8 C 22

9 C 23

10 A 27

11 A 34

12 A 46

shredded 13 B 23

cabbage 14 B 20

(60gx2) 15 B 18

16 C 19

17 C 14

18 C 41

Table 4: Analysis of Variance (2-ways ANOVA) ANOVA

of the mastication amount increase depending on food type

and feedback pattern.

source of sum of degrees of mean F p>F

variance squares freedom squares

Feedback 921 2 460.7 5.18 0.02

pattern

Food 14 1 14.2 0.16 0.69

type

Feedback?Food 21.8 2 10.9 0.12 0.89

Within factors 1066 12 88.9

Total 2024 17

Table 5: Multiple comparison test of the mean mastication

amount increase by each feedback pattern.

compared CI 95% means CI 95% p-value

patterns (low) difference (high)

A B 1.8 16.3 30.8 0.028

A C -0.9 13.7 28.2 0.066

B C -17.2 -2.7 11.9 0.877

HEALTHINF 2019 - 12th International Conference on Health Informatics

446

Table 6: Question contents and corresponding evaluation

items.

Was chewing count correct? accuracy

Did you feel uncomfortable tiresomeness,

with the mounted device? oppressiveness

Wasn’t this system troublesome?

Was the screen easiness to understand

easy to see?

Do you want to use this system practicability

in the future?

Additionally, as a result of subjective evaluation

of speech experiment, utterance awareness increased

for more than half of the subjects in the condition of

using the utterance awareness improvement support

sub-system. Table 6 shows the items corresponding to

the question contents. The results of subjective evalu-

ation on chewing performed in the experiment using

the proposed system are shown in Figure 6. From this

result, it was possible for the subjects to be aware of

mastication during eating without feeling uncomfor-

table with the ear-worn device. Besides, it could be

conﬁrmed that there is no particular problem in the

accuracy and understandability of the proposed sy-

stem. Subsequently, the results of the subjective eva-

luation of speech are shown in Figure 7 and Figure

In this experiment, in order to clarify the diffe-

rence of the awareness with respect to the utterance

Figure 6: Results of subjective evaluation on chewing.

Figure 7: Results of subjective evaluation of feedback by

vibration element.

Figure 8: Results of subjective evaluation of feedback by

LED.

feedback, Fisher’s exact test was carried out for pat-

terns E (vibration) and F (LED). In Fisher’s exact test,

the ﬁve-point scale resulting from subjective evalu-

ation is divided into two categories of ”good” and

”bad”, excluding ”neither” items. ”good” associated

with ”I strongly agree” and, ”I think so,” while ”bad”

with ”I do agree” and ”I strongly disagree.” As a re-

sult, there was a tendency of a signiﬁcant difference in

feedback by pattern E and pattern F (p <0.1) (Table

7). Regarding the feedback of Pattern D, subjective

evaluation was performed only on the item ”Was the

system correctly measuring speech?” (Fig. 9).

Table 7: Subjective evaluation of Fisher’s exact probability

distribution.

Good Bad

vibration 6 1

LED 1 4

Figure 9: Results of subjective evaluation about utterance

accuracy.

5 DISCUSSION

5.1 Accuracy of Mastication and

Utterance Discrimination

As described in the experimental results, the accuracy

of chewing determination in real time was about 91%.

This result improved the accuracy of about 2% when

compared with about 89% of judgment accuracy of

bites count in our previous research (Mitsui et al.,

2017). Also, the accuracy of utterance judgment was

improved compared to previous studies. According

to the result of the subjective evaluation, regarding

the chewing, there were many positive results such as

Effect of Feedback Medium for Real-time Mastication Awareness Increase using Wearable Sensors

447

”I strongly agree” or ”I agree” for the question ”Was

chewing count correct?” Based on this result, it can

be said that chewing was able to be counted without

letting the subject feel uncomfortable. However, the

accuracy of the proposed system may change depen-

ding on the surrounding environment. In this experi-

ment, the accuracy was good because it was carried

out in a quiet room, but there is a drawback that if the

surroundings are noisy, that noise sound is also col-

lected.

As for the utterance, according to the result of the

subjective evaluation, it can be said that subjects were

able to measure without discomfort because accuracy

on Android was mostly positive in subjective evalu-

ation as well. However, in the case of measurement

by Arduino micro-controller, the problem is that the

microphone picks up noise, exceeds the threshold va-

lue, and the code connection method is inappropri-

ate. As a countermeasure, it is conceivable to utilize

the excellent accuracy of the utterance of Android, to

connect Arduino micro-controller to Bluetooth, and

to provide feedback.

5.2 Impact on Meals with or without

Feedback

The number of mastications signiﬁcantly increased by

an average of 16% with real-time feedback. As a

result, it became clear that it is possible to increase

the number of mastications signiﬁcantly through vi-

sual feedback. One of the reasons that the number of

chewing is increasing is that the user becomes more

conscious of chewing during meals due to feedback.

Though the mastication amount increased for all

subjects in all real-time feedback patterns, a signi-

ﬁcantly higher average increase was observed for

feedback pattern A compared to feedback patterns B

and C, between which no signiﬁcant difference was

found. Based on this results, considering the differen-

ces in the device used and interface design between

the three feedback patterns, it is likely that the para-

meters such as a more signiﬁcant amount of informa-

tion and larger screen size inﬂuence more the user to

be conscious of ”chewing.” In the former case, the in-

formation amount due to the feedback in the pattern

A and pattern B is more signiﬁcant in pattern A, and

as a result, the amount of mastication signiﬁcantly

increases more. In the latter case, though pattern A

and pattern C have the same feedback content, feed-

back pattern A that is provided using a smartphone

has a signiﬁcantly higher increase of the mastication

amount than feedback pattern C that is provided using

a smartwatch.

According to the result of the subjective evalua-

tion, the subjects are more conscious of their utte-

rance. Also, regarding Fisher’s exact test, there is

a tendency that pattern E (vibration) is signiﬁcantly

more efﬁcient than pattern F (LED). In the case of

pattern F, if the LED is slightly remove the line of

sight, feedback does not make sense, whereas, in the

case of pattern E, it is easier to force the subject to be

conscious.

6 CONCLUSIONS AND FUTURE

WORKS

As a chewing promotion system usable in everyday

life, we proposed a system that presents feedback of

chewing frequency in real time using bone conduction

microphone and smartphone, and a device to promote

speech. In implementing the proposed system, we

constructed an algorithm for measuring chewing fre-

quency and utterance, conﬁrmed the accuracy of jud-

ging the number of chewing and speech time in the

proposed system, and the inﬂuence of feedback from

the proposed system on meals. As a result, we succee-

ded in judging the chewing frequency with about 91%

judgment accuracy which is about 2% higher than the

previous study. Meanwhile, the duration of utterance

can be measured with an high precision of about 96%.

Besides, the feedback coordinated with the utterance

using Arduino micro-controller controlled wearable

devices enabled to make more than half of the sub-

jects aware of the utterance. Also, there were a diffe-

rence in utterance awareness depending on feedback

medium, showing that vibration stimuli was more ef-

ﬁcient than visual stimuli. However, in the feedback

on utterances, there was a tendency to have negative

responses to the question ”Would you use this system

in the future”? For this reason, it is necessary to deve-

lop a different feedback medium taking into account

the user practicality.

As a prospect, we want to improve the count accu-

racy of mastication in real meal environment (loud

places, various foods) and also consider a feedback

medium with higher visibility and more signiﬁcant ef-

fect to the user. Furthermore, we would like to con-

duct a long-term experiment using the proposed sy-

stem and verify its effectiveness.

REFERENCES

Amft, O., St

ager, M., Lukowicz, P., and Tr

oster, G. (2005).

Analysis of chewing sounds for dietary monitoring. In

UbiComp 2005, 7th Int. Conf. on Ubiquitous Compu-

ting, pages 52–72.

HEALTHINF 2019 - 12th International Conference on Health Informatics

448

Ando, Y., Hanada, N., and Yanagisawa, S. (2008). Does

eating slowly lead to prevent obesity? a literature re-

view. Health Science and Health Care, 8(2):54–63.

Denney-Wilson, E. and Campbell, K. J. (2008). Eating be-

haviour and obesity. BMJ, 337:73–75.

Faudot, T., Lopez, G., and Yamada, I. (2010). Information

system for mapping wearable sensors into healthcare

services: Application to dietary habits monitoring. In

WIVE 2010, 2nd International Workshop on Web In-

tellgence and Virtual Enterprises, pages 1–9.

Fontana, J. and Sazonov, E. (2013). Evaluation of chewing

and swallowing sensors for monitoring ingestive be-

havior. Sensor letters, 11(3):560–565.

Gaul, D., Craighead, W., and Mahoney, M. (1975). Rela-

tionship between eating rates and obesity. Journal of

Consulting and Clinical Psychology, 43(2):123–125.

Inoue, A., Yamasaki, K., and Hoshi, T. (2016). Increasing

the number of mastication by augmented reality. SIG

Technical Reports, (35):1–6. (in Japanese).

Kao (2007). The effect of chewing well, tasting and eating -

preventive measures against metabolic syndrome and

obesity. In Kao Health Care Report, number 19, pages

4–5. (in Japanese).

Kishida, N. and Kamimura, Y. (1993). Relationship of con-

versation during meal and health and dietary life of

school children. The Japanese Journal of Nutririon

and Dietetics, 51(1):23–30.

Kohyama, K., Mioche, L., and Bourdio, P. (2003). Inﬂu-

ence of age and dental status on chewing behavior

studied by emg recordings during consumption of va-

rious food samples. Gerontology, 20(1):15–23.

Kumagai, A., Yanaka, S., Nikai, M., and Kosaka, T. (2016).

Development of a game system using chewing food

for mastication. In EC2016, Entertainment Compu-

ting Symposium, pages 189–190. Information Proces-

sing Society of Japan. (in Japanese).

Li, J., Zhang, N., Hu, L., Ki, Z., Li, R., Li, C., and Wang,

S. (2011). Improvement in chewing activity reduces

energy intake in one meal and modulates plasma gut

hormone concentrations in obese and lean young chi-

nese men. American Journal of Clinical Nutrition,

94(3):709–716.

Logemann, J. (2014). Critical factors in the oral control

needed for chewing and swallowing. Journal of tex-

ture studies, 45(3):173–179.

MHLW (2016). The national health and nutrition survey

in japan, 2014. Japan Ministry of Health Labor and

Welfare. (in Japanese).

Mitsui, H., Nakano, K., Isoyama, N., and Lopez, G. (2017).

Chewing count improvement support system. In DI-

COMO 2017, Multimedia, Distributed, Cooperative,

and Mobile Symposium of the Information Processing

Society of Japan. (in Japanese).

Mizuno, H., Nagai, H., Sasaki, K., Hosaka, H., Sugimoto,

C., and Tatsuta, S. (2007). Wearable sensor system

for human behavior recognition -ﬁrst report: Basic ar-

chitecture and behavior prediction method. In Trans-

ducers 2007, The 14th International Conference on

Solid-State Sensors, Actuators and Microsystems, pa-

ges 435–438.

Nicklas, T., Baranowski, T., Cullen, K., and Berenson,

G. (2001). Eating patterns, dietary quality and obe-

sity. Journal of the American College of Nutrition,

20(6):599–608.

Nishimura, J. and Kuroda, T. (2008). Eating habits mo-

nitoring using wireless wearable in-ear microphone.

In 3rd International Symposium on Wireless Pervasive

Computing, pages 376–381.

Obata, K., Saeki, T., and Tadokoro, Y. (2002). No contact-

type chewing number counting equipment using infra-

red sensor. Transactions of the Society of Instrument

and Control Engineers, 38(9):747–752.

Shuzo, M., Komori, S., Takashima, T., Lopez, G., Tatsuta,

S., Yanagimoto, S., Warisawa, S., Delaunay, J.-J., and

Yamada, I. (2010). Wearable eating habit sensing sy-

stem using internal body sound. Journal of Advan-

ced Mechanical Design Systems and Manufacturing,

4(1):158–166.

Tanigawa, S., Nishihara, H., Kaneda, S., and Haga, H.

(2008). Detecting mastication by using microwave

doppler sensor. In PETRA 2008, 1st international con-

ference on Pervasive Technologies Related to Assistive

Environments, article 88, 7 pages.

Uno, S., Ariizumi, R., and Kaneda, S. (2010). Advising the

number of mastication by using bone-conduction mi-

crophone. In The 24th Annual Conference of the Ja-

panese Society for Artiﬁcial Intelligence, pages 1–4.

Japanese Society for Artiﬁcial Intelligence. (in Japa-

nese).

Zhang, H., Lopez, G., Shuzo, M., and Yamada, I. (2011).

Analysis of eating habits using sound information

from a bone-conduction sensor. In e-Health 2011, Int.

Conf. on e-Health, pages 18–27.

Effect of Feedback Medium for Real-time Mastication Awareness Increase using Wearable Sensors

449