Compressed sensing sparse reconstruction acceleration method for many-core processor

The invention relates to a compressed sensing sparse reconstruction acceleration method for a many-core processor. The method comprises the following steps: acquiring a to-be-sampled signal of to-be-sampled data, and determining an initial sparse vector of the to-be-sampled signal; when compressed sensing is carried out on the to-be-sampled signal, a first measurement matrix and a second measurement matrix are read from the memory, tensor product operation is carried out on the first measurement matrix and the second measurement matrix, a measurement matrix is generated, and the measurement matrix is of a Kronecker structure; and iteratively updating the initial sparse vector according to the generated measurement matrix until a termination threshold condition is met, and obtaining a target sparse vector so as to realize compressed sensing sparse reconstruction of the to-be-sampled signal. By adopting the method, the burden of a memory in the signal processing process can be reduced, the operation efficiency