This invention relates to the field of Markov model technology, specifically to a dietary plan iterative generation
system based on Markov models. The
system receives time-series physiological data collected by the user's
vital signs sensors as an observation sequence, maps a pre-defined food ingredient
library and a
nutrient standard
library into a fixed-length feature hash space, constructs a dimensionality-reduced Hidden Markov
state transition matrix in the feature hash space, compresses high-dimensional food ingredient combination states into low-dimensional hash index states, executes the Viterbi decoding
algorithm in the low-dimensional hash index states to obtain the optimal hidden
state sequence, and restores the low-dimensional hash index sequence in the optimal hidden
state sequence to the food ingredient combination sequence according to the
reverse mapping table corresponding to the hash collision handling strategy, and outputs the dietary plan. This reduces the storage space occupied by the Hidden Markov
state transition matrix, eliminates the risk of memory overflow caused by high-dimensional
state space traversal, and shortens the
response time from the input of the observation sequence to the completion of the decoding of the optimal hidden
state sequence.