classification and training using matlab hmm tool

I am trying to classify a gesture using hmm, I have done the following

1. 1- feature extraction
2. 2- used SOM tool to quantize the features.
3. 3- now i have observation Matrix for the gesture and its repeated gesture, each row is an obseravation vector for each gesture.
4. 4- created the initial transission and emission matrix of the hmm
5. 5- i use hmmtrain(data,transion,emissio) and i have the new transision and emission matricies.
6. 6- when i use hmmcode(data,A,B) error occures.

• attached the input data (all observations)
• the following is the output with the error occuring while running
• after that the code used.

error: " obs_T =

    29
    32
    36
    15
    22
    16
    49
    42
    35
    19
    44
    43
    11
    33
    45
    27
    37
    46
     6
    13
    18
    31
     1
     4
    40
    34
    10
    17
    24
    21
    23
     5
    38
    28

obs_numt =

     1
     5
     6
    17
    19
    20
    33
    34
    49
    65
    69
    71
    78
    80
    97
    99
   102
   106
   113
   114
   116
   127
   129
   131
   145
   146
   161
   163
   170
   193
   241
   257
   273
   289

M =

    34

Error using hmmdecode (line 100) SEQ must consist of integers between 1 and 34.

Error in hmmtrain (line 213) [~,logPseq,fs,bs,scale] = hmmdecode(seq,guessTR,guessE);

Error in test_known_sample_standalone_toask (line 37) [TRANS_EST2, EMIS_EST2] = hmmtrain(cluster_m,Trans_Q,Emis_g)"

here is the code i am using.

% TEST test script for HHM package function test_known_sample_standalone()

clc;
close all;

%observation sequance that will be used.
cluster_m=[3 2 1;4 5 1;2 3 5;4 1 2];

[x y]=size(cluster_m)
O=cluster_m;

%getting all the observation in a the order of its appearance.
 obs_T=unique(O);
[~,index] = unique(O,'first');        %# Capture the index, ignore the actual values
obs_T=O(sort(index))

[~,obs_numt]=ismember(obs_T,O)

%number of distinct observations
M=length(obs_T);

n=length(O); % the number of sequances.
Q=10; %10 states   N states ( Sj)

%Generating Transission matrix NxN
nu = rand(1,Q); nu = nu ./ (sum(nu, 2)*ones(1, Q));

%Generating Transission matrix NxN
Trans_Q = rand(Q,Q); Trans_Q = Trans_Q ./ (sum(Trans_Q, 2)*ones(1, Q));

%generating Emission Matrix NxM
Emis_g = rand(Q, M); Emis_g = Emis_g ./ (sum(Emis_g, 2)*ones(1, M));

%testing for first observation vector [TRANS_EST2, EMIS_EST2] = hmmtrain(cluster_m(1,:),Trans_Q,Emis_g)

PSTATES = hmmdecode(cluster_m,TRANS_EST2, EMIS_EST2)