Computer vision is witnessing a surge of interest in machines accurately recognizing and interpreting human emotions through facial expression analysis. However, variations in image properties such as brightness, contrast, and resolution make it harder for models to predict the underlying emotion accurately. Utilizing a robust architecture of a convolutional neural network (CNN), we designed an efficacious framework for facial emotion recognition that predicts emotions and assigns corresponding probabilities to each fundamental human emotion. Each image is processed with various pre-processing steps before inputting it to the CNN to enhance the visibility and clarity of facial features, enabling the CNN to learn more effectively from the data. As CNNs entail a large amount of data for training, we used a data augmentation technique that helps to enhance the model's generalization capabilities, enabling it to effectively handle previously unseen data. To train the model, we joined the datasets, namely JAFFE and KDEF. We allocated 90% of the data for training, reserving the remaining 10% for testing purposes. The results of the CCN framework demonstrated a peak accuracy of 78.1%, which was achieved with the joint dataset. This accuracy indicated the model's capability to recognize facial emotions with a promising level of performance. Additionally, we developed an application with a graphical user interface for real-time facial emotion classification. This application allows users to classify emotions from still images and live video feeds, making it practical and user-friendly. The real-time application further demonstrates the system's practicality and potential for various real-world applications involving facial emotion analysis.
Citation: Imad Ali, Faisal Ghaffar. Robust CNN for facial emotion recognition and real-time GUI[J]. AIMS Electronics and Electrical Engineering, 2024, 8(2): 227-246. doi: 10.3934/electreng.2024010
[1] |
Jin-Yun Guo, Cong Xiao, Xiaojian Lu .
On |
[2] | Heesung Shin, Jiang Zeng . More bijections for Entringer and Arnold families. Electronic Research Archive, 2021, 29(2): 2167-2185. doi: 10.3934/era.2020111 |
[3] | Fabian Ziltener . Note on coisotropic Floer homology and leafwise fixed points. Electronic Research Archive, 2021, 29(4): 2553-2560. doi: 10.3934/era.2021001 |
[4] | Amira Khelifa, Yacine Halim . Global behavior of P-dimensional difference equations system. Electronic Research Archive, 2021, 29(5): 3121-3139. doi: 10.3934/era.2021029 |
[5] | Tran Hong Thai, Nguyen Anh Dai, Pham Tuan Anh . Global dynamics of some system of second-order difference equations. Electronic Research Archive, 2021, 29(6): 4159-4175. doi: 10.3934/era.2021077 |
[6] | Neşet Deniz Turgay . On the mod p Steenrod algebra and the Leibniz-Hopf algebra. Electronic Research Archive, 2020, 28(2): 951-959. doi: 10.3934/era.2020050 |
[7] | Doston Jumaniyozov, Ivan Kaygorodov, Abror Khudoyberdiyev . The algebraic classification of nilpotent commutative algebras. Electronic Research Archive, 2021, 29(6): 3909-3993. doi: 10.3934/era.2021068 |
[8] | Yusi Fan, Chenrui Yao, Liangyun Chen . Structure of sympathetic Lie superalgebras. Electronic Research Archive, 2021, 29(5): 2945-2957. doi: 10.3934/era.2021020 |
[9] | Dušan D. Repovš, Mikhail V. Zaicev . On existence of PI-exponents of unital algebras. Electronic Research Archive, 2020, 28(2): 853-859. doi: 10.3934/era.2020044 |
[10] | Peigen Cao, Fang Li, Siyang Liu, Jie Pan . A conjecture on cluster automorphisms of cluster algebras. Electronic Research Archive, 2019, 27(0): 1-6. doi: 10.3934/era.2019006 |
Computer vision is witnessing a surge of interest in machines accurately recognizing and interpreting human emotions through facial expression analysis. However, variations in image properties such as brightness, contrast, and resolution make it harder for models to predict the underlying emotion accurately. Utilizing a robust architecture of a convolutional neural network (CNN), we designed an efficacious framework for facial emotion recognition that predicts emotions and assigns corresponding probabilities to each fundamental human emotion. Each image is processed with various pre-processing steps before inputting it to the CNN to enhance the visibility and clarity of facial features, enabling the CNN to learn more effectively from the data. As CNNs entail a large amount of data for training, we used a data augmentation technique that helps to enhance the model's generalization capabilities, enabling it to effectively handle previously unseen data. To train the model, we joined the datasets, namely JAFFE and KDEF. We allocated 90% of the data for training, reserving the remaining 10% for testing purposes. The results of the CCN framework demonstrated a peak accuracy of 78.1%, which was achieved with the joint dataset. This accuracy indicated the model's capability to recognize facial emotions with a promising level of performance. Additionally, we developed an application with a graphical user interface for real-time facial emotion classification. This application allows users to classify emotions from still images and live video feeds, making it practical and user-friendly. The real-time application further demonstrates the system's practicality and potential for various real-world applications involving facial emotion analysis.
Path algebras are very important in representation theory and related mathematical fields, it is interesting to study their counterparts in higher representation theory [21,20]. The
It is well known that the path algebras of acyclic quivers are classified as finite, tame and wild representation types according to their quivers. This classification has a great influence in representation theory. When Iyama and his coauthors study
Regarding
McKay quiver for a finite subgroup of a general linear group is introduced in [27], which connects representation theory, singularity theory and many other mathematical fields. McKay quiver is also very interesting in studying higher representation theory of algebras [20,15]. Over the algebraically closed field of characteristic zero, the preprojective algebras of path algebras of tame type are Morita equivalent to the skew group algebras of finite subgroups of
We also describe the quivers and relations for the tame
The paper is organized as follows. In Section 2, concepts and results needed in this paper are recalled. We recall the constructions of the
In this paper, we assume that
Recall that a bound quiver
Let
A bound quiver
ρ⊥=⋃i,j∈Q0ρ⊥i,j. | (1) |
The quadratic dual quiver of
To define and study
Recall that a bound quiver
Let
The
With an
(Z|n−1Q)0={(u,t)|u∈Q0,t∈Z} |
and arrow set
(Z|n−1Q)1=Z×Q1∪Z×Q1,M, |
with
Z×Q1={(α,t):(i,t)⟶(i′,t)|α:i⟶i′∈Q1,t∈Z}, |
and
Z×Q1,M={(βp,t):(i′,t)⟶(i,t+1)|p∈M,s(p)=i,t(p)=i′}. |
The relation set of
Zρ={∑sas(αs,t)(α′s,t)|∑sasαsα′s∈ρ,t∈Z}, |
ZρM={(βp′,t)(βp,t+1)|βp′βp∈ρM,t∈Z}, |
and
Zρ0={∑s′as′(βp′s′,t)(α′s′,t+1)+∑sbs(αs,t)(βps,t)|∑s′as′βp′s′α′s′+∑sbsαsβps∈ρ0,t∈Z}, |
when the returning arrow quiver
Recall a complete
Given a finite stable
Z⋄˜Q0={(i,n)|i∈Q0,n∈Z}; |
and the arrow set
Z⋄˜Q1={(α,n):(i,n)→(j,n+1)|α:i→j∈Q1,n∈Z}. |
If
ρZ⋄˜Q={ζ[m]|ζ∈˜ρ,m∈Z} | (2) |
here
A connected quiver
Clearly, a nicely graded quiver is acyclic.
Proposition 2.1. 1. Let
2. All the connected components of
3. Each connected component of
Proof. The first and the second assertions follow from Proposition 4.3 and Proposition 4.5 of [12], respectively. The last follows from the definition of
An
Proposition 2.2. Assume that
Proof. Let
ϕ(i,t)=(i,t(n+1)+d(i)−d(i0)). |
It is easy to see that
By definition, an
Proposition 2.3. If
Let
Z⋄Q0[m,l]={(i,t)|m≤t≤l}. |
ρZ⋄˜Q[m,l]={x|x∈ρZ⋄˜Q,s(x),t(x)∈Z⋄Q0[m,l]}. | (3) |
By Proposition 6.2 of [12], we get the following for any
Proposition 2.4.
The complete
Starting with an acyclic
We have the following picture to illustrate the relationship among these quivers.
![]() |
(4) |
The quadratic dual quiver
Recall that a graded algebra
⋯⟶Ptft⟶⋯⟶P1f1⟶P0f0⟶˜Λ0⟶0, | (5) |
such that
Let
The following Proposition justifies the name
Proposition 2.5. Let
Starting with a quadratic acyclic
With the quivers related to
We also associate algebras to the bound quivers
Taking the usual path grading on
So we can get criteria for
Proposition 2.6. Let
Write
![]() |
(6) |
The left triangle is induced by the picture (4) depicting quivers. The left vertically up arrow indicate taking a
Now we assume that
For
At(˜Λ)=(dimke1˜Λte1dimke2˜Λte1⋯dimkem˜Λte1dimke1˜Λte2dimke2˜Λte2⋯dimkem˜Λte2⋯⋯⋯⋯dimke1˜Λtemdimke2˜Λtem⋯dimkem˜Λtem). |
Let
The Loewy matrix
L(˜Λ)=(A1(˜Λ)−E0⋯00A2(˜Λ)0−E⋯00⋅⋅⋅⋯⋅⋅An(˜Λ)00⋯0−EAn+1(˜Λ)00⋯00) | (7) |
with size
Let
l.dimM=(dimMh⋮dimMh+n), | (8) |
where
Assume that
⟶Ps⟶⋯⟶P1→P0⟶M→0 |
such that
l.dimΩsM=Ls(˜Λ)l.dimM | (9) |
by Proposition 1.1 of [18].
Write
V0=(E0⋮0)m(n+1)×m. |
The following Proposition follows from (9) and the definition of the
Proposition 2.7. Assume that
Now assume that
We can restate the Theorems 2.4 and 2.5 of [18] as follows.
Proposition 2.8.
Then by Theorem 3.1 of [30],
The following is part (a) of Proposition 2.9 of [18].
Proposition 2.9. Let
Let
By applying Proposition 2.9, the following is proved as Theorem 2.10 in [18].
Proposition 2.10. If
Recall that for a graded algebra
GKdimΓ=¯limm→∞logmm⨁t=1dimkΓt. | (10) |
By using Koszul duality, the Gelfand-Kirilov dimension of the quadratic dual
The following is Theorem 3.2 in [18]. Let
Theorem 2.11. If
If
The following is Theorem 3.1 in [18].
Theorem 2.12. If
By picture (6), we see an
To classify the
We call an algebra
Lemma 3.1. A stable
Proof. For stable
Now assume that
Clearly, being weakly periodic is a special case of complexity
We have the following characterization of the periodicity and of the complexities for stable
Theorem 3.2. Let
1. The algebra
2. The algebra
3. The algebra
Proof. Assume that
Now assume that
If
Now assume that
Let
Theorem 3.3. Let
1.
2.
3.
Proof. If
If
Combine Theorems 3.2 and 3.3, we get the following.
Theorem 3.4. Let
1.
2.
3.
By Theorem 2.12, we also have the following.
Theorem 3.5. If
Assume that
ΔνΛ=ΔνΓ!,op≃Π(Γ)!,op, |
where
When
Now we define that an
As an immediate consequence of Theorem 3.4, we get a classification of
Theorem 3.6. An
For an
Theorem 3.7. 1.
2.
3.
Proof. This follows from the above Theorems 3.3 and 3.4.
When
Proposition 3.8. If
It is natural to ask if the converse of Proposition 3.8 is true?
By Proposition 3.8, if the
In this section, we assume that
McKay quiver was introduced in [27]. Let
V⊗Si=⨁jSai,jj, 1≤i≤l, |
here
We recall some results on the McKay quivers of Abelian groups and on the relationship between McKay quivers of same group in
Let
G=G(r1,⋯,rm)=Cr1×⋯×Crm |
is a direct product of
(ξi1r1,ξi2r2,⋯,ξimrm)⟶diag(ξi1r1,ξi2r2,⋯,ξimrm), |
for
Proposition 4.1. The McKay quiver
˜Q0(r1,⋯,rm)=Z/r1Z×⋯×Z/rmZ | (11) |
and the arrow set
˜Q1(r1,⋯,rm)={αi,t:i→i+et|i∈Z/r1Z×⋯×Z/rmZ,1≤t≤m}∪{αi,m+1:i→i−e|i∈Z/r1Z×⋯×Z/rmZ}. | (12) |
Proof. We prove by induction on
The Abelian subgroup in
Assume Proposition holds for
Embed the group
(ξi1r1,⋯,ξihrh)⟶diag(ξi1r1,⋯,ξihrh,ξih+1rh+1), |
then group
G(r1,⋯,rh,rh+1)∩SL(h+1,C)=G(r1,⋯,rh), |
and we have
G(r1,⋯,rh,rh+1)=G(r1,⋯,rh)×C′rh+1, |
where
G(r1,⋯,rh,rh+1)/G(r1,⋯,rh,rh+1)∩SL(h+1,C)≃(ξrh+1). |
By Theorem 1.2 of [11], the McKay quiver of
αi,h+1:(i(h),t)→(i(h)−e(h),t+1) |
from one copy
˜Q′1(r1,⋯,rh,rh+1)={αi,t:i→i+e(h+1)t|i∈Z/r1Z×⋯×Z/rh+1Z,1≤t≤h+1} |
as the arrow set.
Now embed
(ξi1r1,⋯,ξih+1rh+1)⟶diag(ξi1r1,⋯,ξih+1rh+1,ξ−i1r1⋯ξ−ih+1rh+1). |
Since Nakayama permutation
This shows that Proposition holds for
Note that the Nakayama permutation for the subgroup of a special linear group is trivial. As a direct consequence of Proposition 3.1 of [11], we also have the following Proposition to describe the McKay quiver of finite group
Proposition 4.2. Let
Let
Theorem 4.3. Let
Let
Note that
Proposition 4.4. Let
Starting with a finite group
Let
Let
{(i,ˉm)|(i,m)∈Z|n−1Q(G),ˉm∈Z/(n+1)Z}. |
For a path
Proposition 4.5.
Proof. For any
⟶˜Ptft⟶˜Pt−1ft−1⟶⋯˜P1f1⟶˜P0⟶˜Λ0(G)ei⟶0 | (13) |
for the simple
Clearly (13) is exact if and only if
(d1,⋯,dht)=(d′1,⋯,d′ht+1)Ct+1. |
Now for any
⋯⟶˜Pt[¯m]˜ft[¯m]⟶˜Pt−1[¯m]˜ft−1[¯m]⟶⋯˜P0[¯m]˜f1[¯m]⟶˜P0⟶Δ(G)ei,¯m⟶0. | (14) |
By comparing the matrices defining the sequences (13) and (14), we see that (13) is exact if and only if (14) is exact. That is, (13) is a projective resolution of simple
Since
By comparing (13) and (14), the simple modules
We call
It is interesting to know if the converse of Proposition 4.5 is true, that is, for an indecomposable nicely graded tame
For the three pairs of quadratic duals of algebras:
For an AS-regular algebra
Theorem 4.6. The following categories are equivalent as triangulate categories:
1. the bounded derived category
2. the bounded derived category
3. the bounded derived category
4. the stable category
5. the stable category
6. the stable category
If the lengths of the oriented circles in
(7) the bounded derived category
(8) the stable category
Proof. By Theorem 4.14 of [28], we have that
By Theorem 1.1 of [7], we have that
By Corollary 1.2 of [7], we have that
On the other hand, we have
By Lemma Ⅱ.2.4 of [19],
Similarly,
When the lengths of the oriented circles in
We remark that the equivalence of (1) and (2) can be regarded as a McKay quiver version of Beilinson correspondence and the equivalence of (1) and (4) can be regarded as a McKay quiver version of Berstein-Gelfand-Gelfand correspondence [5,4]. So we have the following analog to (6), for the equivalences of the triangulate categories in Theorem 4.6:
![]() |
(15) |
In the classical representation theory, we take a slice from the translation quiver and view the path algebra as
Now we consider the case of
Assume that
Since
L(˜Λ)=(M(˜Q)−E0M′(˜Q)0−EE00). | (16) |
This is exactly the Loewy matrix of
Proposition 5.1. Let
1. If there is an arrow from
2. If there is only one arrow from
Proof. The proposition follows directly from
Therefore the number of arrows from
We also have the following Proposition.
Proposition 5.2. If
Proof. Due to that there is no arrow from
Now we determine the relations for the McKay quiver of finite Abelian subgroup of
Let
Let
![]() |
For each vertex
z(γ,i,ci)=ciβi+e1αi+αi+xe2βi,z(β,i,bi)=biαi−eγi+γi+e1αi, |
and
z(α,i,ai)=aiβi−eγi+γi+e2βi, |
for
˜ρcomm(s,r,C(a,b,c))={z(α,i,ai),z(β,i,bi),z(γ,i,ci)|i∈˜Q0}˜ρzero(s,r)={αi+e1αi,βi+e2βi,γi−eγi|i∈˜Q0}, |
and let
˜ρ(s,r,C(a,b,c))=˜ρcomm(s,r,C(a,b,c))∪˜ρzero(s,r). | (17) |
Proposition 5.3. If the quotient algebra
Proof. Assume that
Consider the square with vertices
ciβi+e1αi+c′iαi+e2βi∈I, |
by (2) of Proposition 5.1. Since
z(γ,i,ci)=z(γ,i,ci,1)∈I. |
Similarly, there are
z(β,i,bi)=biαi−eγi+γi+e1αi∈I, |
and
z(α,i,ai)=aiβi−eγi+γi+e2βi∈I. |
For each
αi+e1αi,βi+e2βi,γi−eγi∈I, |
by Proposition 5.2.
So
Let
ei˜Λ2=kβi−e2αi−e+kγi+eβi+be1+kγi+eαi+be2, |
and
˜Λ3ei=kγi+eβi+e1αi, |
by computing directly using the relations in
dimkei˜Λ3ei≤1=dimkei˜Λ3(G(s,r))ei, |
and
dimkei˜Λ2ei′{≤1,for i′=i−e,i+e1,i+e2;=0,otherwise. |
This implies that
dimkei˜Λ2ei′≤dimkei˜Λ2ei′, |
for any
So
ei′˜Λtei=ei′˜Λt(G(s,r))ei |
for all
This proves
For the quadratic dual quiver
Proposition 5.4. If the quotient algebra
˜ρ⊥(s,r,C(a,b,c))={z(α,i,−a−1i),z(β,i,−b−1i),z(γ,i,−c−1i)|i∈˜Q0}. | (18) |
Proof. Let
Now construct the quiver
Recall by (2), the relation set for
z(γ,(i,t),ci)=ciβi+e1,t+αi+e2,t+1βi,t,z(β,(i,t),bi)=biαi−e,t+1γi,t+γi+e1,t+1αi,t, |
and
z(α,(i,t),ai)=aiβi−e,t+1γi,t+γi+e2,t+1βi,t, |
for
˜ρcomm(s,r,C(a,b,c))[t]={z(α,(i,t),ai),z(β,(i,t),bi),z(γ,(i,t),ci)|i∈˜Q0}˜ρzero(s,r)[t]={αi+e1,t+1αi,t,βi+e2,t+1βi,t,γi−e,t+1γi,t|i∈˜Q0}, |
and let
ρZ⋄˜Q(s,r)=⋃t∈Z(˜ρcomm(s,r,C(a,b,c))[t]∪˜ρzero(s,r)[t]). | (19) |
By taking a connected component
![]() |
Here we denote by
Let
ρ(s,r)={x∈ρZ⋄˜Q(s,r)|e[0,2]xe[0,2]=x}=˜ρcomm(s,r,C(a,b,c))[0]∪˜ρzero(s,r)[0]. |
Since any sequence of relations in
ρ(s,r)={αi+e1,1αi,0,βi+e2,1βi,0,γi−e1,1γi,0|i∈Z/sZ×Z/rZ}∪{αi+e2,1βi,0−βi+e1,1αi,0,αi−e,1γi,0−γi+e1,1αi,0,αi−e1,1γi,0−γi+e1,1αi,0|i∈Z/sZ×Z/rZ}. | (20) |
Proposition 5.5. If the quotient algebra
The quadratic dual quiver
ρ⊥(s,r)={αi+e2,d+1βi,d1=βi+e1,d+1αi,d,αi−e,1γi,0+γi+e1,1αi,0,αi−e1,1γi,0+γi+e1,1αi,0|i∈Z/sZ×Z/rZ}. | (21) |
Proposition 5.6. If the quotient algebra
So the relations for the
Let
![]() |
![]() |
The vertices for
We have immediate the following on the arrows of these McKay quivers.
Lemma 5.7. Let
1. There is a loop at each vertex of
2. There is at most one arrow from
3. There is an arrow
4. There are at most
Let
Lemma 5.8. If
αj,hαi,j,βj,hβi,j,αj,hβi,j,βj,hαi,j∈I(Ξ) |
if such paths exist.
Proof. If
αj,hαi,j∈I(Ξ). |
Similarly we have
βj,hβi,j,αj,hβi,j,βj,hαi,j∈I(Ξ) |
if such paths exist.
Write
˜ρ11(Ξ)={αj,hαi,j|i,h,j∈˜Q0(Ξ),i≠h},˜ρ12(Ξ)={αj,hβi,j|i,h,j∈˜Q0(Ξ),i≠h},˜ρ21(Ξ)={αj,hβi,j|i,h,j∈˜Q0(Ξ),i≠h},˜ρ22(Ξ)={βj,hβi,j|i,h,j∈˜Q0(Ξ),i≠h}. |
Take
˜ρp(Ξ)=˜ρ11(Ξ)∪˜ρ12(Ξ)∪˜ρ21(Ξ)∪˜ρ22(Ξ). | (22) |
As a corollary of Lemma 5.8, we get the following.
Proposition 5.9.
Let
˜ρa(Ξ,Ca)={αi,jγi−γjαi,j,βj,iγj−γiβj,i|αi,j∈˜Q1,ai,j∈Ca,i<j}. | (23) |
Proposition 5.10. By choosing the representatives of the arrows suitably, we have a set
˜ρa(Ξ,Ca)⊆I(Ξ). |
Proof. By Lemma 5.7, there is an arrow
βj,iγj−bj,iγiβj,i∈I(Ξ), |
for some
αi,jγi−a′i,jγjαi,j∈I(Ξ), |
for some
Starting from
βj,iγj−γiβj,i∈I(Ξ) |
for all arrows
This proves that by choosing the representatives of the arrows suitably, we have
Write
Lemma 5.11. For each arrow
Write
˜ρa⊥(Ξ,Ca)={ai,jαi,jγi+γjαi,j,βj,iγj+γiβj,i|αi,j∈˜Q1,ai,j∈Ca,i<j}. | (24) |
Now consider
Fix
μi,j={αi,ji<j,βi,ji>j, and ζj,i={βj,ii<j,αj,ii>j. | (25) |
Then
Consider the following cases.
Lemma 5.12. Assume that there is only one arrow
1. If
2. We have
{γ2i−ciζj,iμi,j,ciγ2i+ζj,iμi,j} | (26) |
is a orthogonal basis for
Proof. Apply Proposition 5.1 for the arrow
The second assertion follows from direct computation.
Lemma 5.13. Assume that there are exactly two arrows
1. There is a
2. If
3. We have
{biζj1,iμi,j1+ζj2,iμi,j2,ζj1,iμi,j1−b′iζj2,iμi,j2,γ2i}and{biζj1,iμi,j1−ζj2,iμi,j2,ciζj1,iμi,j1−γ2i,ζj1,iμi,j1+biζj2,iμi,j2+ciγ2i} | (27) |
are orthogonal bases for
Proof. By Proposition 5.1, we have that the images of
If
The rest follows from direct computations.
Lemma 5.14. Assume that there are three arrows
1. There are
biζj1,iμi,j1+ζj2,iμi,j2,b′iζj1,iμi,j1+ζj3,iμi,j3∈I(Ξ). |
2. If
3. We have
{biζj1,iμi,j1−ζj2,iμi,j2,b′iζj1,iμi,j1−ζj3,iμi,j3,γ2i−ciζj1,iμi,j1,ciγ2i+ζj1,iμi,j1+biζj2,iμi,j2+b′iζj3,iμi,j3}and{γ2,biζj1,iμi,j1−ζj2,iμi,j2,b′iζj1,iμi,j1−ζj3,iμi,j3,ζj1,iμi,j1+biζj2,iμi,j2+b′iζj3,iμi,j3,γ2} | (28) |
are orthogonal bases of
Proof. The lemma follows from Proposition 5.1, similar to above two lemmas.
Denote by
Ci={{ci}⊂k∗i∈˜Q01(Ξ),{ci,bi}⊂k∗i∈˜Q02(Ξ),{ci,bi,b′i}⊂k∗i∈˜Q03(Ξ), | (29) |
and set
C′i={∅i∈˜Q01(Ξ),{bi}⊂k∗i∈˜Q02(Ξ),{bi,b′i}⊂k∗i∈˜Q03(Ξ). | (30) |
Let
Ui(Ci)={{ciγ2i+ζj,iμi,j}i∈˜Q01(Ξ),{ζj1,iμi,j1−biζj2,iμi,j2,γ2i−ciζj1,iμi,j1}i∈˜Q02(Ξ),{biζj1,iμi,j1−ζj2,iμi,j2,b′iζj1,iμi,j1−ζj3,iμi,j3,γ2i−ciζj1,iμi,j1}i∈˜Q03(Ξ). | (31) |
and let
U−i(C′i)={{ζj,iμi,j}i∈˜Q01(Ξ),{γ2,ζj1,iμi,j1−biζj2,iμi,j2}i∈˜Q02(Ξ),{γ2i,biζj1,iμi,j1−ζj2,iμi,j2,b′iζj1,iμi,j1−ζj3,iμi,j3}i∈˜Q03(Ξ). | (32) |
Lemma 5.15. A basis of the orthogonal subspace of
U⊥i(Ci)={{ζj,iμi,j+ciγ2i}i∈˜Q01(Ξ),{ζj1,iμi,j1+b−1iζj2,iμi,j2+c−1iγ2i}i∈˜Q02(Ξ),{ζj1,iμi,j1+b−1iζj2,iμi,j2+b′−1iζj3,iμi,j3+ciγ2i}i∈˜Q03(Ξ). | (33) |
and let
U−,⊥i(C′i)={{ζj,iμi,j}i∈˜Q01(Ξ),{biζj1,iμi,j1+ζj2,iμi,j2}i∈˜Q02(Ξ),{ζj1,iμi,j1+b−1iζj2,iμi,j2+b′−1iζj3,iμi,j3}i∈˜Q03(Ξ). | (34) |
Proof. This follow immediately from (2) of Lemma 5.12, (3) of Lemma 5.13 and (3) of Lemma 5.14.
Fix
CJ=Ca∪⋃i∈Q0∖JCi∪⋃i∈JC′i, | (35) |
for
˜ρ(Ξ,J,CJ)=˜ρp(Ξ)∪˜ρa(Ξ,Ca)∪⋃i∈˜Q0∖JUi(Ci)∪⋃i∈JU−i(C′i). | (36) |
By Lemma 5.11 and Lemma 5.15, we have the following.
Proposition 5.16.
˜ρ⊥(Ξ,J,CJ)=˜ρa⊥(Ξ,Ca)∪⋃i∈˜Q0∖JU⊥i(Ci)∪⋃i∈JU−,⊥i(Ci). | (37) |
Proposition 5.17. Let
Proof. Take
Let
It is also immediate that
By Lemma 5.16, a quadratic dual relation of
Proposition 5.18. Let
Now construct the nicely-graded quiver
ρZ⋄˜Q(Ξ)=ρZ⋄˜Q(Ξ)(J,CJ)={z[m]|z∈˜ρ(Ξ,J,CJ),t∈Z} |
for some parameter set
\Lambda(\Xi,J,C) \simeq k {\mathbb Z}_{ \diamond} \widetilde{Q}(\Xi)/(\rho_{ {\mathbb Z}_{ \diamond} \widetilde{Q}(\Xi)}(J,C_J)) , |
if
By taking the complete
![]() |
![]() |
They are all nicely-graded quivers. We get
\begin{equation} \label{{relnpg}}{\rho(\Xi,J,C) = \{z[0]| z\in \tilde{\rho}(\Xi, J, C)\}.} \end{equation} | (38) |
For any parameter set
Write
\mu_{i,j,t} = \left\{\begin{array}{ll} \alpha_{i,j,t} & i < j, t = 0,1,\\ \beta_{i,j,t} & i > j, t = 0,1, \end{array}\right. \mbox{ and } \zeta_{j,i} = \left\{\begin{array}{ll} \beta_{j,i,t} & i < j,t = 0,1,\\ \alpha_{j,i,t} & i > j, t = 0,1. \end{array}\right. |
Let
\begin{equation} U_i(C_i) = \left\{ \begin{array}{ll}\{ \gamma_{i,1} \gamma_{i,0} +\zeta_{j,i,1}\mu_{i,j,0}\} & i\in \widetilde{Q}_{01}(\Xi), \\ \{ \zeta_{j_1,i,1}\mu_{i,j_1,0 } - \zeta_{j_2,i,1}\mu_{i,j_2,0 }, \gamma_{i,1} \gamma_{i,0} - \zeta_{j_1,i,1}\mu_{i,j_1,0}\} & i\in \widetilde{Q}_{02}(\Xi), \\ \{ \zeta_{j_1,i,1}\mu_{i,j_1,0 } - \zeta_{j_2,i,1}\mu_{i,j_2,0 }, \\ \quad \zeta_{j_1,i,1}\mu_{i,j_1,0 } - \zeta_{j_3,i,1}\mu_{i,j_3,0 }, \gamma_{i,1} \gamma_{i,0} - \zeta_{j_1,i,1}\mu_{i,j_1,0} \} & i\in \widetilde{Q}_{03}(\Xi), \end{array}\right. \end{equation} | (39) |
and let
\begin{equation} U^-_i(C'_i) = \left\{ \begin{array}{ll}\{ \zeta_{j,i,1}\mu_{i,j,0} \} & i\in \widetilde{Q}_{01}(\Xi), \\ \{ \gamma^2, \zeta_{j_1,i,1}\mu_{i,j_1,0 } - \zeta_{j_2,i,1}\mu_{i,j_2,0 } \} & i\in \widetilde{Q}_{02}(\Xi), \\ \{ \zeta_{j_1,i,1}\mu_{i,j_1,0 } + \zeta_{j_2,i,1}\mu_{i,j_2,0 }+ \zeta_{j_3,i,1}\mu_{i,j_3,0 } \} & i\in \widetilde{Q}_{03}(\Xi), \end{array}\right. \end{equation} | (40) |
\begin{equation} U_i^{\perp}(C_i) = \left\{ \begin{array}{ll}\{ \zeta_{j,i,1}\mu_{i,j,0} + \gamma_{i,1} \gamma_{i,0} \} & i\in \widetilde{Q}_{01}(\Xi), \\ \{ \zeta_{j_1,i,1}\mu_{i,j_1,0 } + \zeta_{j_2,i,1}\mu_{i,j_2,0 }+ \gamma_{i,1} \gamma_{i,0} \} & i\in \widetilde{Q}_{02}(\Xi), \\ \{ \zeta_{j_1,i,1}\mu_{i,j_1,0 } + b^{-1}_i\zeta_{j_2,i,1}\mu_{i,j_2,0 }+ \\ \quad \zeta_{j_3,i,1}\mu_{i,j_3,0 } + \gamma_{i,1} \gamma_{i,0} \} & i\in \widetilde{Q}_{03}(\Xi), \end{array}\right. \end{equation} | (41) |
and let
\begin{equation} U^{-,\perp}_i(C'_i) = \left\{ \begin{array}{ll}\{ \zeta_{j,i,1}\mu_{i,j,0} \} & i\in \widetilde{Q}_{01}(\Xi), \\ \{ \zeta_{j_1,i,1} \mu_{i,j_1,0 } + \zeta_{j_2,i,1}\mu_{i,j_2,0 } \} & i\in \widetilde{Q}_{02}(\Xi), \\ \{ \zeta_{j_1,i,1}\mu_{i,j_1,0 } + \zeta_{j_2,i,1}\mu_{i,j_2,0 }+ \zeta_{j_3,i,1}\mu_{i,j_3,0 } \} & i\in \widetilde{Q}_{03}(\Xi). \end{array}\right. \end{equation} | (42) |
Take
\begin{array}{rl}\rho_p(\Xi) = & \{ \alpha_{i,j,1} \alpha_{j,h,0} | i,h ,j \in \widetilde{Q}_0(\Xi), i\neq h \} \cup \{ \beta_{i,j,1} \beta_{j,h,0} | i,h ,j \in \widetilde{Q}_0(\Xi), i\neq h \} \\ &\cup \{ \beta_{i,j,1} \alpha_{j,h,0}| i,h ,j \in \widetilde{Q}_0(\Xi), i\neq h \} \\ &\cup \{ \beta_{i,j,1} \alpha_{j,h,0} | i,h ,j \in \widetilde{Q}_0(\Xi), i\neq h\} ,\end{array} |
and
\rho_a(\Xi) = \{ \alpha_{i,j,1} \gamma_{i,0} - \gamma_{j,1} \alpha_{i,j,0}, \beta_{j,i,1} \gamma_{j,0} - \gamma_{i,1} \beta_{j,i,0} | \alpha_{i,j}\in \widetilde{Q}_1, i < j \}. |
Write
{\rho_a}^\perp(\Xi) = \{ \alpha_{i,j,1} \gamma_{i,0} + \gamma_{j,1} \alpha_{i,j,0}, \beta_{j,i,1} \gamma_{j,0} + \gamma_{i,1} \beta_{j,i,0} | \alpha_{i,j}\in \widetilde{Q}_1, i < j\}. |
They are subsets of the space
Take a subset
\begin{equation} \rho(\Xi,J) = \rho_p(\Xi) \cup \rho_a(\Xi) \cup \bigcup\limits_{i\in \widetilde{Q}_0\setminus J} U_{i,0} \cup \bigcup\limits_{i\in J} U^-_{i,0}, \end{equation} | (43) |
and
\begin{equation} \rho^{\perp}(\Xi,J) = {\rho_a}^\perp(\Xi) \cup \bigcup\limits_{i\in \widetilde{Q}_0\setminus J} U^{\perp}_{i,0} \cup \bigcup\limits_{i\in J} U^{-,\perp}_{i,0}. \end{equation} | (44) |
We have the following descriptions of the relation sets for the
Proposition 5.19. Let
Proposition 5.20. Let
1. By constructing
2. The complete
3. Though we need the field
We would like to thank the referees for reading the manuscript carefully and for suggestions and comments on revising and improving the paper. They also thank the referee for bring [24] to their attention.
[1] |
Albornoz EM, Milone DH, Rufiner HL (2011) Spoken emotion recognition using hierarchical classifiers. Comput Speech Lang 25: 556‒570. https://doi.org/10.1016/j.csl.2010.10.001 doi: 10.1016/j.csl.2010.10.001
![]() |
[2] |
Erol BA, Majumdar A, Benavidez P, Rad P, Choo KKR, Jamshidi M (2019) Toward artificial emotional intelligence for cooperative social human-machine interaction. IEEE Transactions on Computational Social Systems 7: 234‒246. https://doi.org/10.1109/tcss.2019.2922593 doi: 10.1109/tcss.2019.2922593
![]() |
[3] |
Cohn JF, Ambadar Z, Ekman P (2007) Observer-based measurement of facial expression with the Facial Action Coding System. The Handbook of Emotion Elicitation and Assessment 1: 203‒221. https://doi.org/10.1093/oso/9780195169157.003.0014 doi: 10.1093/oso/9780195169157.003.0014
![]() |
[4] |
Vaillant R, Monrocq C, Le Cun Y (1994) Original approach for the localization of objects in images. IEE Proceedings-Vision, Image and Signal Processing 141: 245‒250. https://doi.org/10.1049/ip-vis:19941301 doi: 10.1049/ip-vis:19941301
![]() |
[5] |
Rowley HA, Baluja S, Kanade T (1998) Neural network-based face detection. IEEE T Pattern Anal 20: 23‒38. https://doi.org/10.1109/34.655647 doi: 10.1109/34.655647
![]() |
[6] | Jain V, Learned-Miller E (2010) FDDB: A benchmark for face detection in unconstrained settings. Technical Report UMCS-2010-009, University of Massachusetts, Amherst. Available from: https://people.cs.umass.edu/~elm/papers/fddb.pdf |
[7] |
Zhu X, Ramanan D (2012) Face detection, pose estimation, and landmark localization in the wild. IEEE Conference on Computer Vision and Pattern Recognition, 2879‒2886. https://doi.org/10.1109/cvpr.2012.6248014 doi: 10.1109/cvpr.2012.6248014
![]() |
[8] |
Yan J, Zhang X, Lei Z, Li SZ (2014) Face detection by structural models. Image Vision Comput 32: 790‒799. https://doi.org/10.1109/fg.2013.6553703 doi: 10.1109/fg.2013.6553703
![]() |
[9] |
Viola P, Jones M (2001) Rapid object detection using a boosted cascade of simple features. Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 1‒9. https://doi.org/10.1109/cvpr.2001.990517 doi: 10.1109/cvpr.2001.990517
![]() |
[10] |
Lienhart R, Maydt J (2002) An extended set of Haar-like features for rapid object detection. Proceedings of the IEEE International Conference on Image Processing, 1‒4. https://doi.org/10.1109/icip.2002.1038171 doi: 10.1109/icip.2002.1038171
![]() |
[11] |
Belhumeur PN, Hespanha JP, Kriegman DJ (1997) Eigenfaces vs. Fisherfaces: Recognition using class-specific linear projection. IEEE T Pattern Anal 19: 711‒720. https://doi.org/10.1109/34.598228 doi: 10.1109/34.598228
![]() |
[12] |
Yang MH, Kriegman DJ, Ahuja N (2002) Detecting faces in images: A survey. IEEE T Pattern Anal 24: 34‒58. https://doi.org/10.1109/34.982883 doi: 10.1109/34.982883
![]() |
[13] |
Wu J, Zhang C, Xue T, Freeman B, Tenenbaum J (2016) Learning a probabilistic latent space of object shapes via 3D generative-adversarial modeling. Advances in Neural Information Processing Systems, 82‒90. https://doi.org/10.1609/aaai.v32i1.12223 doi: 10.1609/aaai.v32i1.12223
![]() |
[14] |
Sroubek F, Milanfar P (2011) Robust multichannel blind deconvolution via fast alternating minimization. IEEE T Image Process 21: 1687‒1700. https://doi.org/10.1109/tip.2011.2175740 doi: 10.1109/tip.2011.2175740
![]() |
[15] |
Krizhevsky A, Sutskever I, Hinton GE (2012) ImageNet classification with deep convolutional neural networks. Advances in Neural Information Processing Systems, 1097‒1105. https://doi.org/10.1145/3065386 doi: 10.1145/3065386
![]() |
[16] | Sermanet P, Eigen D, Zhang X, Mathieu M, Fergus R, LeCun Y (2013) Overfeat: Integrated recognition, localization, and detection using convolutional networks. arXiv preprint arXiv: 1312.6229. https://doi.org/10.48550/arXiv.1312.6229 |
[17] |
Bartlett MS, Littlewort G, Frank M, Lainscsek C, Fasel I, Movellan J (2006) Fully automatic facial action recognition in spontaneous behavior. 7th International IEEE Conference on Automatic Face and Gesture Recognition, 223‒230. https://doi.org/10.1109/fgr.2006.55 doi: 10.1109/fgr.2006.55
![]() |
[18] |
Pantic M, Rothkrantz LJ (2004) Facial action recognition for facial expression analysis from static face images. IEEE T Syst Man Cy B 34: 1449‒1461. https://doi.org/10.1109/tsmcb.2004.825931 doi: 10.1109/tsmcb.2004.825931
![]() |
[19] |
Tian YI, Kanade T, Cohn JF (2001) Recognizing action units for facial expression analysis. IEEE T Pattern Anal 23: 97‒115. https://doi.org/10.1109/cvpr.2000.855832 doi: 10.1109/cvpr.2000.855832
![]() |
[20] |
Ng HW, Nguyen VD, Vonikakis V, Winkler S (2015) Deep learning for emotion recognition on small datasets using transfer learning. Proceedings of the 2015 ACM on International Conference on Multimodal Interaction, 443‒449. https://doi.org/10.1145/2818346.2830593 doi: 10.1145/2818346.2830593
![]() |
[21] |
Chaudhari A, Bhatt C, Krishna A, Travieso-González CM (2023) Facial emotion recognition with inter-modality-attention-transformer-based self-supervised learning. Electronics 12: 1‒15. https://doi.org/10.3390/electronics12020288 doi: 10.3390/electronics12020288
![]() |
[22] |
Yang D, Huang S, Wang S, Liu Y, Zhai P, Su L, et al. (2022) Emotion recognition for multiple context awareness. Proceedings of the European Conference on Computer Vision, 144‒162. https://doi.org/10.1007/978-3-031-19836-6_9 doi: 10.1007/978-3-031-19836-6_9
![]() |
[23] |
Song C, Ji S (2022) Face Recognition Method Based on Siamese Networks Under Non-Restricted Conditions. IEEE Access 10: 40432‒40444. https://doi.org/10.1109/access.2022.3167143 doi: 10.1109/access.2022.3167143
![]() |
[24] |
Qu X, Zou Z, Su X, Zhou P, Wei W, Wen S, et al. (2021) Attend to where and when: Cascaded attention network for facial expression recognition. IEEE Transactions on Emerging Topics in Computational Intelligence 6: 580‒592. https://doi.org/10.1109/tetci.2021.3070713 doi: 10.1109/tetci.2021.3070713
![]() |
[25] |
King DE (2009) Dlib-ml: A machine learning toolkit. The Journal of Machine Learning Research 10: 1755‒1758. https://doi.org/10.1145/1577069.1755843 doi: 10.1145/1577069.1755843
![]() |
[26] | Lyons MJ, Kamachi M, Gyoba J (1997) Japanese female facial expressions (JAFFE). Database of Digital Images. https://doi.org/10.5281/zenodo.3451524 |
[27] |
Goeleven E, De Raedt R, Leyman L, Verschuere B (2008) The Karolinska directed emotional faces: a validation study. Cognition and Emotion 22: 1094‒1118. https://doi.org/10.1080/02699930701626582 doi: 10.1080/02699930701626582
![]() |
1. | Jin Yun Guo, Yanping Hu, On n-hereditary algebras and n-slice algebras, 2024, 00218693, 10.1016/j.jalgebra.2024.07.020 | |
2. | Jin Yun Guo, Yanping Hu, Deren Luo, Multi-layer quivers and higher slice algebras, 2024, 23, 0219-4988, 10.1142/S021949882450186X |