diagram - 维特比解码器 k=4

Question

在 k=3 格子图中，我使用此代码将我的距离相加并选择最短的

tmpPrevState=[]
    for a in range(4): 
        if tmpDist[0, 2 * a + 0] <= tmpDist[0, 2 * a + 1]:
            cumDist[a] = tmpDist[0, 2 * a + 0]
            tmpPrevState.append((a % 2) * 2 + 0)
        else:
            cumDist[a] = tmpDist[0, 2 * a + 1]
            tmpPrevState.append((a % 2) * 2 + 1)
    prevState.append(tmpPrevState)

我将范围设置在 0~3 之间，因为 K=3 状态图中有 4 个状态，但在 k=4 环境中，图中有 8 个状态，所以我将范围更改为 0~7，但这个解码器不能正常工作

在此之下，是我的整个编解码器代码

 import numpy as np

def Encoder(data):
data = np.append(data, [0, 0, 0, 0])#shift Register
dataSize = np.shape(data)[0] #(64,)
shiftReg = [0, 0, 0, 0] #k=4
encoded_bit = np.zeros((2,dataSize)

for i in range(dataSize):
    shiftReg[3] = shiftReg[2]
    shiftReg[2] = shiftReg[1]
    shiftReg[1] = shiftReg[0]
    shiftReg[0] = data[i]
    encoded_bit[0, i] = np.logical_xor(np.logical_xor(shiftReg[0], shiftReg[1]), np.logical_xor(shiftReg[2], shiftReg[3]))
    encoded_bit[1, i] = np.logical_xor(np.logical_xor(shiftReg[0],shiftReg[2]), shiftReg[3])
return encoded_bit

def ViterbiDecoder(encoded_bit):
ref_out = np.zeros((2,16))
ref_out[0, :] = [0, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 0, 1]
ref_out[1, :] = [0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 0, 1, 0, 1, 1, 0]
dataSize = np.shape(encoded_bit)[1] 
cumDist = [0, 100, 100, 100, 100, 100, 100, 100] 

prevState = []
for i in range(dataSize):
    tmpData = np.tile(encoded_bit[:, i].reshape(2,1),(1,16))
    dist = np.sum(np.abs(tmpData - ref_out), axis=0) 
    tmpDist = np.tile(cumDist, (1,2)) + dist

    tmpPrevState=[]
    for a in range(8): 
        if tmpDist[0, 2 * a + 0] <= tmpDist[0, 2 * a + 1]:
            cumDist[a] = tmpDist[0, 2 * a + 0]
            tmpPrevState.append((a % 2) * 2 + 0)
        else:
            cumDist[a] = tmpDist[0, 2 * a + 1]
            tmpPrevState.append((a % 2) * 2 + 1)
    prevState.append(tmpPrevState)
    state_index = np.argmin(cumDist)
   

decoded_bit=[]
for b in range(dataSize-1,-1,-1):
    decoded_bit.append(int(state_index/2))
    state_index = prevState[b][state_index]
data_size = np.shape(decoded_bit)[0]
decoded_bit = np.flip(decoded_bit)[0:data_size - 4]
return decoded_bit