I am trying to evaluate my processed speech for my DSP project. My project is an implementation of an MMSE algorithm proposed by Y.Ephraim and D. Malah, "Speech Enhancement Using a Minimum Mean Square Error Short-Time Spectral Amplitude Estimator," in the IEEE Transactions on Acoustics, Speech, and Signal Processing, Vol. ASSP-32, No. 6, Dec. 1984. As per project requirements, I ran PESQ, LLR, and segment SNR. I got numbers, but I am not sure what they mean. Is high number is good or bad? Also, what does it mean if I get a negative number?

Please help and thanks in advance


1 Answer 1


Well, I submitted my project. Thanks to the help of this group.

As for the PESQ number, I think the range is between 1 and 5, where 5 being excellent and 1 is low.

Anyhow, I did another trick to estimate my performance. Instead of feeding the functions (SESQ, LLR, and Segment SNR) the clean speech and the processed speech, I fed them clean speech with the clean speech. I got numbers and these were my reference (or maximum) numbers. Then I compared all my runs against these reference numbers


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.