Noise Suppression comparison at Mobile World Congress 2013
This year, the booth of NXP Software at Mobile World Congress 2013 was almost entirely devoted to voice enhancements i.e. features like noise suppression and echo cancellation required to enhance the quality of voice calls. One of the demos at their booth was a simple A-B pairwise listening test, in which several competing 2-microphone noise suppression solutions are compared against each other.
The items in the listening test are recordings made from commercially available solutions. The phones under test are the Huawei Ascend D1, the Sony Xperia T, the Apple iPhone5 and two different tunings of LifeVibes VoiceExperience. The listening test is blind i.e. the booth visitor gets to hear two audio clips A and B but he does not know to which phone A and B correspond. The assignment of A and B to the 5 different solutions is randomized for every listener.
Two different noise conditions have been used: babble noise, which contains both stationary and non-stationary components and a music track which is highly non-stationary. The noise is played at two different levels.
After having played audio clip A and B, the question for the visitor to answer is simple: Which solution do you prefer, A or B, or no preference.
After four days of testing at Mobile World Congress, 613 votes have been casted. From these votes, 82 times a visitor indicated ‘equal’. Of the remaining 531 votes, the graph below indicates how many times the listeners preferred each solution against any of the other 4 solutions.
On average, LifeVibes VoiceExperience is clearly more often preferred over the other 4 solutions in the test, followed by the Huawei Ascend D1, the iPhone 5 and the Sony Xperia T.